AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained...

151
AN ARCHITECTURAL COMPARISON OF CONTEMPORARY APPROACHES AND PRODUCTS FOR INTEGRATING HETEROGENEOUS INFORMATION SYSTEMS Principal Investigator: Dr. Amar Gupta Technical Advisor: Prof. Stuart E. Madnick Researchers: Teresa Wingfield and Christopher Poulsen Working Paper 3084-89 November 1989 Comments and suggestions should be directed to: Dr. Amar Gupta Principal Research Associate Sloan School of Management MIT Room E40-265 Cambridge, MA 02139 Telephone: (617) 253-8906 0 Massachusetts Institute of Technology, 1989. All rights reserved. Contents of this working paper cannot be photocopied or quoted without advance written approval from Principal Investigator.

Transcript of AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained...

Page 1: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

AN ARCHITECTURAL COMPARISON OF CONTEMPORARYAPPROACHES AND PRODUCTS FOR INTEGRATING

HETEROGENEOUS INFORMATION SYSTEMS

Principal Investigator: Dr. Amar GuptaTechnical Advisor: Prof. Stuart E. MadnickResearchers: Teresa Wingfield and

Christopher Poulsen

Working Paper 3084-89November 1989

Comments and suggestions should be directed to:

Dr. Amar GuptaPrincipal Research AssociateSloan School of ManagementMIT Room E40-265Cambridge, MA 02139Telephone: (617) 253-8906

0 Massachusetts Institute of Technology, 1989. All rightsreserved. Contents of this working paper cannot bephotocopied or quoted without advance written approval fromPrincipal Investigator.

Page 2: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

TABLE OF CONTENTS

ABSTRACT

1.01.1.1.2.1.3.1.4.1.5.1.6.1.7.

INTRODUCTIONBackground..............Functional Deficiencies.Levels of Integration...Constraints..............Models and Schemata.....Languages ...............General Directions......

..................... .1-1..................... 1-2..................... 1-4

..................... 1-7. . ................... 1-8.................... 1-12

.................... 1-13

SECTION 2.0 HIGHLIGHTS OF PRINCIPAL PRODUCTS AND APPROACHES2.1. BBN Systems and Technologies: Cronus ........2-22.2. Booz Allen: IDIMS ............ 2-4..........2-42.3. Cambridge Technology Group: SURROUND ........2-62.4. Cincom: SUPRA . .............................. 2-82.5. Cognos Corporation: POWERHOUSE ............. 2-102.6. Control Data Corporation: ASCENT . .......... 2-122.7. Control Data Corporation: CDM*PLUS .........2-152.8. Data Integration, Inc.: Mermaid ............2-172.9. Digital Equipment Corp.: Rdb/VMS, DBMS ..... 2-202.10.Gupta Technologies, Inc.: SQL System ....... 2-222.11.Honeywell Federal Systems,

Database Testbed System....2.12.IBM: DB2, SQL/DS, SAA, and2.13.Information Builders, Inc.:

Inc.:

othersFOCNE

Distribut..····...

Te..e..es

2.14.Interbase Software Corporation: INTERBA2.15.Massachusetts Institute of Technology:

Composite Information Systems..........2.16.Metaphor Computer Systems, Inc.: DIS...2.17.0n-Line Software International, Inc.:

RAMIS Information System .... .........2.18.Oracle Corporation: ORACLE.............2.19.Relational Technology, Inc.: INGRES....2.20.Software AG: ADABAS....................2.21.Sybase, Inc.: SYBASE...................2.22.Tandem Computers Incorporated: NonStop2.23.TRW: TDIE ..............................2.24.Xerox Advanced

Multibase.....Information Technology:

....·.......·. ·..........

SE.

ed.2-24.2-27.2-30.2-32

.... 2-34

.... 2-38

...

. .

·. ·

SQL, . .

.2-40

.2-42

.2-44

.2-46

.2-48

.2-5i

.2-52

DISCUSSION AND CONCLUSIONUnintrusive Integration ..............Intrusive Integration..................Interfacing with Controlled Redundancy.Interfacing with Transaction Transfer..Additional Evaluation Criteria.........Conclusion.............................

...... 3-1

...... 3-3

...... 3-4

...... 3-5

...... 3-6

...... 3-9

REFERENCES .............. R-1, R-2

BIBLIOGRAPHY ..... ..... ....... ............. B-1, B-2

SECTION

SECTION 3.0.3.1.3.2.3.3.3.4.3.5.3.6.

Page 3: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

ABSTRACT

Virtually all large organizations are faced with the problem

of intelligently accessing and integrating information maintained

in islands of disparate computer-based systems. Since a number of

vendors claim that their respective products and services can

overcome the problem of integration, a comprehensive study was made

into the technical characteristics of all these products and

services, as well as additional ones identified through other

means. This report documents the results of this study in terms

of a new taxonomy for classifying various approaches for

integrating heterogeneous information systems. In addition, the

principal concepts and issues pertaining to this field are also

presented in this report.

Page 4: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

SECTION ONE: INTRODUCTION

1.1. Background

Most organizations today depend on a portfolio of information

processing machines, ranging from mainframes to microcomputers and

from general purpose workstations to very specialized systems, to

meet their computational requirements. While the individual

hardware and software components in the growing array of computer

systems may meet the objectives for which each was initially

designed, their heterogeneity presents a major obstacle to ready

access and assimilation of information. It requires tremendous

amounts of time and effort to retrieve information from multiple

systems today, and in many situations, it is simply not feasible

to obtain integrated responses in desired timeframes.

As compared to the intra-organizational situation discussed above,

the issues involved in integrating information on

interorganizational basis are even more complex. First, there is

greater heterogeneity in technical capabilities across

organizations as compared to a single organization. Second, there

is less likelihood of reaching consensus on common goals and

standards. Third, non-technical issues, such as access rights and

sharing of costs, acquire added importance when multiple

organizations are involved. Integration of relevant information

from various computer-based systems is indeed a challenging task.

1-1

Page 5: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

1.2. Functional Deficiencies

The problem of inefficient, incomplete, and time-consuming access

to information can be traced, from a technical viewpoint, to

functional deficiencies at several levels as described below:

Structured and Unstructured Applications

Conventional and computer-based information systems have

been designed with specific applications in view. These

systems are efficient for performing the originally

intended application, but they are inefficient in dealing

with ad hoc queries or new unanticipated applications.

Information versus Knowledge

Traditional database systems focus on retrieval of data,

and on performing elementary operations (e.g., sorting

and merging). Such systems cannot use both data and

programs to respond intelligently to queries like "How

many vehicles can be made available to evacuate personnel

from the vicinity of a disaster site within a specified

period of time?" which require analysis of many factors.

Diverse types of Information

Different types of information (numerical, graphic,

1-2

Page 6: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

pictorial, speech, and video) are referenced in very

different ways. During the sixties, computers were

designed to manipulate, store, and retrieve numerical

information. During the seventies, the focus was on

textual information. Relatively less effort has so far

been directed towards efficient storage and retrieval of

pictorial information and its combination with numerical

and textual information.

Communications

In spite of continuing effort by a number of national

and international institutions, it is still difficult to

transfer complex information across computers of

dissimilar architectures.

Granularity

It is difficult to judge what volume of information

should be made available, and how it should be arranged

and tailored to meet the needs of the person requesting

the information.

Security

It is necessary to determine who can access the

information, and what each person is permitted to modify.

1-3

Page 7: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

III

This aspect is especially important for applications

involving more than one organization.

Semantics

It is difficult to specify exactly the subset of

information desired without knowing details of the

systems being accessed. Further, a particular piece of

information may possess different connotations on various

systems. Conversely, the same item (such as a particular

spare part) may be specified by different numbers on

various systems. The above list describes a few of the

key technical problems.

In addition, there are major non-technical impediments that

restrict integration. Because of the large number of issues and

variables, a global solution to the problem of integrating

distributed information systems has remained an elusive dream.

Partial solutions are becoming available, however, and these

solutions are identified and analyzed in later sections of this

report.

1.3. Levels of Integration

Integration of information resources can be visualized at different

levels. At the least complex end of the spectrum, information can

be exchanged across a network using basic facilities such as remote

1-4

Page 8: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

procedure call [13] and message passing. At the next higher level,

a new layer of software can be created to insulate users from

idiosyncrasies of different host machines. This layer can provide

uniformity at the level of logically centralized file servers as

in the case of Andrew at CMU [2], or at the level of application

programs as in the case of Project Athena at MIT [3].

The approaches described in the preceding paragraph require

significant degree of user involvement. Also, they represent

instances of integration of information. The focus of this report

is, however, on integration of information systems. Further, the

emphasis is on systems "dissimilar" from each other in terms of:

*Hardware, for example, one information system using anIBM mainframe and a second system functioning onUnisys hardware;

*Operating systems, for example, one information systemusing OS/MVS, and other using UNIX;

*Data models, for example, one system using networkdata models, and the other using relational datamodels;

*Database management systems, for example, one usingINGRES, and the other using ORACLE databasemanagement systems, respectively.

Apart from the above differences arising from dissimilarities in

hardware and software, different information systems use different

sets of design decisions and underlying assumptions. For example,

one system may be using annual salary of personnel while another

may be using monthly salary; further, the salary in one case may

be gross salary (typical of budgeting systems), whereas in the

second case, it may be net salary (typical of systems used for

printing checks). Similarly, a particular spare part may be

1-5

Page 9: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

ILI

specified by one part number in one system and by a different part

number in a second system. All these differences must be

reconciled when attempting to integrate dissimilar information

systems.

There are two facets to integration--physical integration and

logical integration. Physical integration, accomplished through

physical connectivity, refers to the process of establishing actual

communication links among disparate systems. Logical integration

refers to the process of accessing disparate systems in concert

for generating integrated responses. This involves reconciliation

of different assumptions and perspectives embedded in the systems

being integrated. At the logical level, the integration system

must:

*Know where all the information is stored, along withthe data formats and the query languages of thelocal systems;

*Decompose the query into subqueries that can beexecuted by local systems;

*Accumulate the results from all the subqueries;

*Reconcile differences among the results accumulated;and

*Formulate integrated answers.

All the above tasks must ideally be carried out with no

modification to existing systems. However, many products discussed

later in this report require significant changes to existing

information systems.

Based on the level of integration achieved, and some additional

1-6

Page 10: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

characteristics, various types of integration approaches can be

distinguished. These approaches, using the terminology of

different researchers who have used dissimilar terms to refer to

same or similar ideas, include the following:

*Cooperation System for Heterogeneous Data BaseManagement System [4].

*Federated Database [5,9].

*Composite Information System [6].

*Knowledge Based Integrated Information System [7,11].

*Multidatabase [8,10].

*Superdatabase [12].

Instead of attempting to cover all the views expressed by the

individuals who coined these terms and phrases, the endeavor in

this report is to encapsulate the main concepts and the emerging

technologies and products that enable integration of dissimilar

information systems.

1.4. Constraints

The design of an integrated information system is constrained by

the following facts:

(a) The new system must integrate existing information

systems, each of which was designed independently with

little premonition, if any, that the system would form

part of a larger system at a future date.

(b) Each existing information system must be fully understood

in order to comprehend all underlying assumptions and

semantics. Rarely are these systems fully and accurately

1-7

Page 11: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

II'

documented.

(c) The design must allow flexibility to add or remove

individual systems into the integrated configuration. As

such, it requires detailed knowledge of both immediate

and long-term constituents of the integrated system.

(d) The design must involve the least amount of changes to

existing systems. Transactions, irrespective of whether

they are queries or updates, involve retrieval or

manipulation of information maintained in individual

systems, each of which imposes its own set of

requirements and standards which must be reconciled by

the new system.

The goal is to mitigate all the above-mentioned problems in an

efficient and effective manner.

1.5. Models and Schemata

A Distributed Heterogeneous Database Management System attempts to

meet the above goal by providing the facilities to access, to

aggregate, and to update information maintained in multiple,

distributed, heterogeneous systems. A typical distributed

heterogeneous approach is characterized by the following:

a) The use of a common data model, with appropriate mapping

to data models of participating systems;

(b) The support of a standard user language for retrieving

data on a global basis in all cases, and also for storing

data in some cases;

1-8

Page 12: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

(c) The ability to decompose a user request into an

equivalent set of subqueries expressed in the format and

manner required by participating systems;

(d) The capability to resolve incompatibilities in semantics,

data structures, and other parameters;

(e) *The mechanism to take the individual pieces of

information provided by participating systems and to

integrate them in order to provide a cohesive response

to the original user request.

In addition to the above, some approaches facilitate implementation

of sophisticated features such as authorization control and the

ability to draw inferences in situations involving missing or

conflicting data.

When data need to be retrieved from a single file, it is sufficient

to simply know the file design including the keys for accessing

records. One single view or schema of the file design meets the

requirements of the systems analyst and the user. However, when

there are several disparate information systems and there are many

users from different disciplines, each user tends to look at the

information from his or her perspective. By having more than one

schema, it is feasible to handle complex problems, as well as to

achieve data independence.

Most experts recognize the requirement for three types of schema.

This notion of a three-schema architecture originated in 1977, when

the Study Planning and Requirements Committee (SPARC) of the

1-9

Page 13: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

III

American National Standards Institute (ANSI) published a report on

this subject [14]. The three types of schema are as follows:

(a) Global conceptual schema: This schema is intended as a

blueprint that shows the existence of all the information

maintained in all the information systems. To enable

intelligent access and integration, the conceptual schema

must be designed to incorporate the semantics of all

stored data. For this reason, it is sometimes referred

to as logical schema. There is only one global schema

for the entire enterprise.

(b) Local or internal schema: This schema covers the data

in an individual information system, that may itself be

hosted on one or more computers. Since there is one

internal schema for each information system, a complex

environment necessarily implies a large number of such

local data models. The collection of such local models

is referred to as a local schemata.

(c) External or application schema: This schema describes

the user or the application view of the distributed data.

Since each class of users or applications will view the

data in a different manner, more than one external schema

is required. This collection is referred to as external

schemata.

The strategy used to map information between these three types of

schemata determines the overall capability to mediate across

dissimilar architectures, to maintain integrity of data, and to

meet user expectations. Transformation of information between the

1-10

Page 14: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

three schemata is ideally managed via a three-schema data

dictionary. Use of very powerful data definition language is

critical in defining the global schema in terms of objects, events,

and integrity constraints [15].

It is important to emphasize that many commercial products and

approaches do not support the concept of a global schema. Instead,

a relational table or matrix is established between the required

external schemata and the available internal schemata. Such a two-

schema approach, referred to as the interfacing approach, offers

no technological control over data integrity [15]. Data are

exchanged either through transaction passing or by controlled

redundancy. In the former case, all transactions are transmitted

from the source database to the destination database, where they

are transformed to adhere to the requirements of the external

schemata. In controlled redundancy, instantaneous copies of the

source database are stored at multiple nodes of the network, and

these copies are replaced by new versions at periodic intervals.

Both alternatives of the interfacing approach attempt to meet new

user and application needs in an ad hoc manner, without any

consistent set of rules governing the overall computational

environment.

To overcome the above problems, it is necessary to use a three-

schema approach, especially in applications of a critical or a

strategic nature. All the major research programs funded by the

U.S. Government, such as Multibase, Integrated Information Support

1-11

Page 15: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

III

System, and Integrated Manufacturing Distributed Database

Administration System, adopted the three-schema approach.

Comparative evaluation of eight major initiatives in this area

appears in [16,17]. To maintain adequate integrity of shared data

of critical and strategic importance, the need for utilizing a

three-schema approach cannot be overemphasized. This aspect

acquires added importance when an application needs to access data

from various DBMS-based information systems, as well as from

earlier file systems that contain no notion of DBMS.

1.6. Languages

The only database language that has gained endorsement from major

computer manufacturers, and from standards establishing

organizations such as ANSI and ISO, is SQL. SQL statements can be

used to create tabular views that approximate the notion of an

external schema. Further, it can be used to create a conceptual

schema, although a semantically weak one [18]. Intended to be used

for queries, SQL has no internal schema.

There are several flavors of SQL in existence. A number of vendors

claim to offer support for multiple versions of SQL. Further, some

SQL based integration approaches can deal with legacy systems that

offer no support for SQL. In such cases user queries expressed in

SQL are converted into the native languages of the participating

information systems.

1-12

Page 16: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

There is a recent trend towards using graphics techniques to select

databases, files, and tables. From a user perspective, these icon-

oriented approaches are far friendlier than traditional language

oriented techniques. However, there is still no common graphics

interface that has been endorsed by a critical mass of vendors.

Even with a proprietary graphics interface, navigation with icons

may be especially relevant for users at the top echelons of the

organization.

1.7. General Direction

The basic concepts outlined above have been used in evaluating the

principal products and approaches that are commercially sold. The

list of vendors contacted is placed at Appendix A, and detailed

specifications are tabulated in Appendix B. For the subset of

products and approaches that hold some degree of relevance, a brief

write-up was generated on each product. These write-ups are

presented in Section 2 of this report. In Section 3, a new

taxonomy is created for classifying different approaches to

integration, and recommendations are made for selecting an

appropriate set of technologies.

1-13

Page 17: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

III

Page 18: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

SECTION TWO: HIGHLIGHTS OF PRINCIPAL PRODUCTS AND APPROACHES

In Section 1, the fundamental principles of integrating dissimilar

information systems were presented. In this section, principal

products and approaches are discussed and analyzed using these

fundamental principles.

The products and approaches described in this section were derived

from numerous sources, including articles in computer journals and

magazines, proceedings of relevant conferences, lists of conference

participants, and personal discussions with representatives in

various companies. From the initial list generated, the systems

previously covered in References [1] and [16] were deleted, unless

the complexion of the system had changed significantly during the

intervening period. The material presented in this section is not

an exhaustive survey of all issues, but rather a discussion that

covers the whole field taking all significant design approaches in

view. It should be emphasized that some companies offer customized

services, rather than a product, making it difficult to evaluate

them in detail. In this section, each product is described by an

introduction, its functional architecture, and its salient

features. Additional details are tabulated in Appendix B.

2-1

Page 19: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

2.1. BBN Systems and Technologies: Cronus

Cronus is an environment designed to support coherent integration

of heterogeneous computer systems. The objectives of Cronus are:

to provide an effective, general-purpose, distributed computing

environment composed of hosts with differing hardware and operating

systems; to support an integrated computing environment consisting

of both the native operating system environment and the Cronus

environment; and to provide comprehensive support for development

of large-scale distributed applications. Cronus began as a

research project funded by the Rome Air Development Center and the

Naval Ocean Systems Center in 1981. The project was a two-year

joint effort between BBN Systems and Technologies Corporation and

Xerox Advanced Information Technology, both of Cambridge, Mass.

The first prototype from this project became operational in 1984.

Cronus supports both object-oriented and relational local data

models. Global updates may be performed on an object by object

basis. Update operators for insertion, deletion, and modification

are in algebraic form and operate associatively on sets of tuples.

Stored update procedures may include conditional statements and

expressions.

FUNCTIONAL ARCHITECTURE

Cronus runs on top of existing operating systems on various

hardware platforms. The Global Schema definition is supported for

object-oriented data storage and retrieval. Every system resource

2-2

Page 20: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

is an object and is accessed through operations defined by the type

of the object. Cronus supports heterogeneity by serving as a

bypassable layer of abstraction between applications programs and

native operating systems. Through this approach, both conventional

access to native operating systems resources and services and

access to a coherent, uniform (object-oriented) system interface,

regardless of computer system base, are supported.

Distributed query processing and update capabilities require

translation of high-level requests into sequences of algebraic

operations on data stored at possibly different sites. The

algebraic approach defines a collection of "physical" operators on

tuples and takes into account the sort order, the presence of

indexes, and the physical location. Transformation rules determine

when two algebraic expressions are equivalent, and optimization

rules specify when it is profitable to transform an expression into

another equivalent one.

Salient Features

Cronus incorporates an evolutionary approach in its design. Cronus

claims that its Open Architecture approach makes it possible to

upgrade or replace components or technology with little or no

impact on the rest of the system. There is uniformity across

diverse operating systems, hardware platforms, and applications.

2-3

Page 21: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

III

2.2. Booz Allen: IDIMS

IDIMS is an expert system oriented approach designed by Booz Allen

to provide automated data access. The user obtains access

throughout the distributed data environment with a single global

manipulation language and a user-friendly, intelligent interface.

Upon receipt of a query, IDIMS decomposes the query into a set of

database calls and function applications; determines security

violations on databases; formulates queries in appropriate

languages and queries databases; applies functions to retrieve data

and finally displays the data via a state of the art graphic

interface, in a style appropriate to the type and the number of

data elements to be displayed.

It is important to note that IDIMS is not a commercial "off the

shelf" product, but instead a conceptual framework used by Booz

Allen to achieve automated data access. Although a complete

production system has not yet been implemented, Booz Allen claims

that it has built systems which relate to each functional component

of IDIMS.

FUNCTIONAL ARCHITECTURE

The IDIMS system consists of a Query Processor, a Knowledge base

for database models and function models, a Query Language

Generator, and a Natural Language Processor.

The Query Processor translates a global data manipulation language

2-4

Page 22: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

query into a series of data and function calls using the database

and function models knowledge bases. In doing so, it plans the

appropriate databases and functions to apply, chooses between plans

using query optimization and time consideration techniques, and

executes plans by calling the Query Language Generator to retrieve

data. In addition, it determines whether there are potential

security violations for the local databases queried.

The Data Base Models Knowledge Base is a structure which maintains

data models for each of the distributed local databases. These

models contain information and rules on security consideration,

representative query language, data, data type, and host computer.

The Function Models Knowledge Base is a structure which maintains

knowledge about the available procedures to alter or determine data

attributes. Each function model provides for a specific logical

data linking procedure such as correlate, sum, and join, and

describes domain, range, and process definition. The Query

Processor uses the Function Models Knowledge Base to determine

which functions satisfy given constraints.

The Query Language Generator translates the global manipulation

language into the appropriate local database query language. It

then makes these queries and retrieves the data. Accordingly, the

Query Language Generator maintains translations for each local

database language and procedures to send queries to each computer

on the network.

2-5

Page 23: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

III

The Natural Language Processor provides an additional interface

through which individuals can use limited natural language

capabilities to query databases. This processor then translates

the natural language into the IDIMS global manipulation language

using conceptual dependency and case grammar.

SALIENT FEATURES

IDIMS is not a product, but an approach that eliminates the need

for the user to know multiple query languages, locations of data,

database paradigms, and manipulation schemata. A high-level,

iconic, global query language is supported; however, updates are

not supported.

2.3. Cambridge Technology Group: SURROUND

The Cambridge Technology Group (CTG) of Cambridge, Mass.,

specializes in providing customized distributed heterogeneous data

integration within short time frames. A small group of users and

managers from the client company interface with SURROUND developers

at CTG to implement software interfaces as well as a rapid

prototype of a functional application within a one week period.

The goal of CTG is to construct a user-friendly system which

focuses on one particular strategic- application.

CTG applications support queries and updates. Local data models

are supported via SURROUND interface and coordination tools.

Complete data transparency is provided for geographically dispersed

2-6

Page 24: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

locations via dialup ASCII modems.

FUNCTIONAL ARCHITECTURE

The UNIX operating system is used as a major part of the SURROUND

service to resolve incompatibilities of different systems.

Software tools access various distributed environments, extract

information from individual programs, and then transfer data to

the UNIX environment where the information is joined together.

Talk Tools are communication tools which imitate the code that the

host computer understands. The host computer accepts commands from

the Talk Tool and sends results back to the same Talk Tool in the

UNIX environment. There are also a variety of tools for attaining

connectivity, security, network management, and for gauging

performance.

The construction of the global schema is tailored to the needs of

the application. SURROUND does not define a data manipulation

language; it supports the interface currently used by the user.

SQL formatted files are accessed with modified versions of SQL

Data Handler and communication tools on remote system perform

global updates. Commander, a proprietary tool, handles transaction

management functions.

SALIENT FEATURES

Rapid prototype systems have been developed for TRANSCOM, General

Electric, and other organizations within one week spans of time.

2-7

Page 25: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

CTG claims that it can provide production software in as little as

three months. Record structures can be changed easily and new

databases can be added quickly, making the SURROUND environment

flexible.

2.4. Cincom: SUPRA

SUPRA, a product of Cincom, provides a commercial implementation

of the ANSI/SPARC Three Schema Architecture. Such an architecture,

described in Section 1.5, insulates applications from changes in

the physical storage and conceptual database, reduces development

and maintenance costs, and increases the useful life of

applications.

FUNCTIONAL ARCHITECTURE

The SUPRA system integrates distributed heterogeneous databases

and manages distributed processing by having a SUPRA kernel reside

locally within each host machine or database. This has the effect

of making the hosts aware of the integrated database network and

in essence creates a "standard protocol" through which the

databases can communicate.

The SUPRA kernel consists of two basic engine components, the

Distributed Relational Data Manager (DRDM), and the Heterogeneous

Data Management Processor (HDMP), along with a portion of the

Global Directory. Using the kernel on each local host, the

requesting node coordinates the entire unit of work, including the

2-8

Page 26: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

decomposition of the query into separate neutral protocol requests

and the determination of where to send them. The local/receiving

nodes, each with its own kernel, translate the request into the

appropriate local data manipulation language, gather the data,

perform calculations, and return the result to the requesting node.

In this decentralized structure, each node is intelligent and has

the ability to coordinate with each other via knowledge of the

network in the resident Global Directory. This approach to

distributed database and processing management assumes that most

of the data accessed by any given node reside locally and that an

"independent" node, with the ability to process locally, has

significant speed and reliability advantages. In those instances

where the data are not local, the Global Directory contains the

information to tell the DRDM where to get them. As such, the

network is not dependent on the reliability of a central Global

Directory or engine. If any one node goes down, only the data

residing at that node become inaccessible, not the whole system.

Additionally, by distributing the necessary processing information,

every transaction need not access a central directory, thus

eliminating a potential bottleneck. However, with more than one

copy of the Global Directory, the SUPRA approach implies more

maintenance and overhead compared to a single copy approach. Every

time the Global Directory is updated at one node, the update must

be reflected across the entire network creating a "cascade" of

transactions throughout the network.

2-9

Page 27: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

III

SALIENT FEATURES

SUPRA currently provides for the use of synonyms and the resolution

of scale conflicts. The decentralized approach allows a synonym

table oriented to each user's particular application and working

language. The Global Directory contains the ability to retain

rules for performing scale translation and calculation before data

integration. A unique feature called "triggers", alerts the user

to the possibility of a data conflict. For example, a trigger

could be established to ensure that two numbers, before they are

combined together, be within a certain percentage of each other.

At execution time, when a user wishes to combine the two numbers,

a violation of this requirement would lead to an interrupt, and the

user to be prompted about the problem.

SUPRA utilizes a two phase commit procedure which provides good

concurrency control. While assuring secure updates for SUPRA or

other database management systems which support two phase commits,

the product currently only allows updates from join views to

distributed databases which use SQL.

2.5. Cognos Corporation: POWERHOUSE

PowerHouse consists of four components: PowerHouse 4GL, PowerHouse

StarBase, StarNet and StarGate. PowerHouse 4GL is a fourth-

generation language; PowerHouse StarBase is a relational

database management system; StarNet is network routing software;

and StarGate is a gateway provider to other supported local

2-10

Page 28: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

database management systems. The PowerHouse 4GL was developed by

Cognos Corporation, Ottowa, Ontario while the remaining components

were developed by InterBase Software Corporation, Bedford, MA.

Cognos began to market the Relational DBMS of InterBase and its

distributed technology in conjunction with the PowerHouse

application development language in 1988.

FUNCTIONAL ARCHITECTURE

In this approach, there is no concept of a single global schema.

Instead, multiple relational schemata may be referenced, even from

within the same program. Relational operators are utilized to

selectively access and merge pieces of data. Research on a

distributed query engine is currently in progress.

A discussion of the capabilities of InterBase components is

provided later under its own prototype overview. The main

difference between PowerHouse and InterBase is global data

manipulation language supported. Both products support SQL and

Groton Data Manipulation Language, but the stand-alone PowerHouse

4GL is unique to PowerHouse.

One of the benefits of the PowerHouse 4GL is that it provides

access to relative Record Management Service files and network

databases in addition to sequential and indexed Record Management

Service files and relational databases supported by InterBase.

The Open Systems Relational Interface (OSRI) is a relational

database protocol. PowerHouse 4GL elements for transaction

2-11

Page 29: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

III

processing, report generation and batch updating use a relational

access approach which makes it possible to embed OSRI calls into

the elements. This means that as more local data models are added

to the configuration, it is possible to develop applications which

access them using PowerHouse.

The PowerHouse 4GL product set includes PhD master data dictionary.

Using PhD, it is possible to treat all meta-data as if they were

resident in a single logical dictionary. Thus, one need not

redefine meta-data in the PhD dictionary.

SALIENT FEATURES

The salient features listed in the description of the InterBase

prototype also apply here. Cognos is attempting to incorporate

expert systems and object-oriented techniques in the PowerHouse

product.

2.6. Control Data Corporation: ASCENT

ASCENT is derived from the Technology Information System (TIS)

developed by Lawrence Livermore National Labs (LLNL). Work on this

system originally commenced in the early seventies. CDC has a

technology transfer license with LLNL (actually with the University

of California) to support this technology, and to take it into

production.

ASCENT is a suite of software products that provides

2-12

Page 30: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

communications, networking, office automation, and database

services within an interactive environment that is common to all

users. The *gateway portion of ASCENT supports a powerful

interpretive scripting language that can simulate an interactive

user when connected to a heterogeneous remote computer. This

gateway serves as a tool that can be programmed with the scripting

language to log into databases and to extract data from a broad

range of remote systems.

FUNCTIONAL ARCHI TECTURE

A collection of software modules, known as the Intelligent Gateway

Processor (IGP), use a UNIX system as the link between menu type

front-end systems and geographically and technically disparate

mainframes. The IGP front-end program, Integrated Information

System (IIS), offers menus composed of descriptor files which might

consist of classes of information resources, specific connections

to remote hosts, local database options, or available system

utilities. In addition to menus, IIS understands a command set

that overlaps with the UNIX Shell command set. As IIS is

implemented in a script-driven interpreter, UNIX commands can be

recognized even without arguments, and prompts can be enabled for

arguments or options.

The Network Access Machine (NAM), an extended implementation of

the specifications published by the National Institute of Standards

and Technology, utilizes a variety of communication methods to

provide access to other computer systems. Currently NAM can

2-13

Page 31: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

III

interface with direct asynchronous connections between two hosts,

most autodial modems, TCP/IP networks and X.24 networks. NAM uses

an interpreted communication language to specify the logic

necessary to establish connections and conduct dialogues with

remote systems. This language includes structured constructs,

extensive pattern matching capabilities and capabilities to insert

"filters" in an existing connection. These filters are specialized

programs written to modify the appearance of the data.

In a heterogeneous environment, ASCENT*gateway must be driven by

the "host UNIX system" database management system to extract

specific data elements from their remote computers under script

language control. The DHDBMS places these data elements into the

proper place in the "logical record" requested by the user. This

approach requires no software to be added to any of the remote

systems since each system feels that the scripting language

interactions with other remote systems are identical in appearance

to interactions with its own users.

SALIENT FEATURES

ASCENT*gateway allows automated access and connectivity to local

or remote computing resources. ASCENT*mail provides complete

electronic mail and message services within a computer network

including addressing, forwarding, and mail file management

functions. A menu-driven user interface is available to access

integrated software applications, utilities, and system

administration tools.

2-14

Page 32: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

2.7. Control Data Corporation: CDM*PLUS

The concept of CDM*plus was developed and prototyped with direction

and funding from the Integrated Information Support System (IISS,

pronounced I2S 2) initiative of the U.S. Air Force. CDM*plus defines

and categorizes data using the ANSI/SPARC Three Schema Architecture

discussed in Section 1.5.

FUNCTIONAL ARCHITECTURE

The CDM*plus dictionary/directory is loaded and maintained via the

Neutral Data Definition Language (NDDL) which describes all meta-

data to the ANSI/SPARC based three schema structure. NDDL is SQL

based and describes objects with verbs such as create, define,

alter, and drop. Each schema has unique commands. For example,

in the internal schema, the NDDL can describe the three basic types

of database models. NDDL provides inter-schema mappings using the

conceptual schema or the data model as the base. NDDL is a human

and machine readable language that can be used in either in

interactive mode or in batch mode, as well as on forms.

The mapping strategy between the schema allows data to be accessed

without regard to the location or the storage mechanism. For

example, an application can join two relational user views from the

external schema that each map to two entities. These entities in

turn can be mapped to multiple databases and database types on

multiple computers without burdening the user or developer with all

of the details of the actual conversions, translations or mappings.

2-15

Page 33: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

III

This independence is critical to provide a buffer for life cycle

maintenance as applications and storage structures change, and to

provide the capability for full relational operation (including

outer joins and referential integrity) between heterogeneous

databases.

Queries can be embedded into current applications written in Cobol,

C, or Fortran. These queries are precompiled by CDM*plus using the

external schema view and transformed into actual database queries.

All code necessary to access the databases is generated including

DBMS precompilations. Since the base schema is the conceptual

schema or data model, domain constraint checking and referential

integrity checking can be automatically enforced based on the

conceptual schema. After the code is generated, it is transferred

by the Code Manager to the appropriate computers where local

compiles, DBMS precompiles, and links are accomplished.

All query processing is controlled by two CDM*plus components:

the Network Transaction Manager (NTM), which controls unique

process to process communication capabilities that are necessary

in the distributed environment, and the CDM*plus Distributed

Request Supervisor (DRS), which controls the actual queries on each

host and formulates the messages sent to the NTM. This approach

allows serial or parallel processing of queries. In the serial

process, one query produces input for other queries.

Each local node contains an NTM and multiple CDM*plus runtime

(CDMR) modules to facilitate distributed processing. The CDMR

2-16

Page 34: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

incorporates DRS, Aggregators, File Utilities and Code Manager.

Query modules in applications that attempt to integrate information

from multiple heterogeneous databases can be replaced by NDML

statements. This allows for easier maintenance of legacy

applications because changes to the systems only require re-

precompilation to generate all the necessary code.

SALIENT FEATURES

CDM*plus requires the user to model his or her enterprise and to

load that model into the CDM. The three schema architecture

provides one global conceptual view of the complete data. In

addition, it provides independence between the actual physical

storage of the data, the users' view of the data, and the

conceptual view of the data. Modeling and integration of

applications and databases with CDM*plus is done in an evolutionary

manner, allowing for the implementation of one application at a

time rather than a full rewrite of the entire system.

2.8. Data Integration, Inc.: Mermaid

Mermaid, a trademark of Unisys, is an integrated data access

product marketed by Data. Integration, Inc. It allows users of

multiple relational or network database management systems, running

on different machines, to access and manipulate data using standard

SQL, even when the participating systems don not support SQL.

Integration may require translation of the data into a standard

data type, translation of the units, combination or division of

2-17

Page 35: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

II1

fields, and encoding values. Only updates to a single database are

presently supported. Updates to replicated and fragmented

relations may cross databases, but the updates to all databases are

not made concurrently.

The Mermaid system is designed for an environment in which there

are many databases that are independently administered and where

it is not feasible to standardize on one DBMS. With Mermaid, new

applications may be developed either at the global level when

distributed access is required, or above an existing old database.

Local administrators grant access permissions to users of Mermaid

and then Mermaid adds an additional level of access control.

FUNCTIONAL ARCHITECTURE

The major processes are: User Interface Process, Distributor

Process, DBMS Driver Process, and Data Dictionary/Directory (DD/D).

The User Interface Process contains the SQL parser, the query

validator, the query library, and the report generator. The

Distributor Process contains the optimizer and the controller. The

optimizer plans query execution and the controller starts processes

and coordinates execution. There is one DBMS Driver Process for

each database to be accessed. It contains the schema translation,

the language translation, and the DBMS interface. The Data

Dictionary/Directory (DD/D) is a database that contains information

about the schemata, users, host computers, and network. It

supports three layers of schema definition.

2-18

Page 36: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

A user input is either a command to execute within the user

interface or an SQL query. The SQL query is parsed, validated,

translated into Distributed Intermediate Language (DIL), and sent

to the distributor process which performs the Distributed Execution

Monitor function. The optimizer plans the execution by consulting

the DD/D for the location of the relations. It sends commands and

subqueries to the controller for transmission to the DBMS drivers.

The driver reads a message which may be a command to initialize,

terminate, execute schema and data of the query from the global

form into the local form and then translates it from DIL into the

local DBMS language. The first time that a relation is retrieved,

the data are translated into the global form.

SALIENT FEATURES

Mermaid treats files as large objects. Qualification is done on

structured data. The user selects qualified objects to view. The

system then starts a retrieval process and a display process that

are customized for the file type.

Communication between processes is done using standard protocols.

The main processes use TCP/IP on an Ethernet. If TCP/IP is not

available on the computer with the DBMS, another protocol may be

used to communicate between the controller and the driver or

between the driver and the DBMS.

Mermaid is written as a set of processes with well defined

interfaces so that it can be customized for different environments.

2-19

Page 37: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

III

It can also be embedded in or called by other products and it can

use other products to process the output.

2.9. Digital Equipment Corp.: Rdb/VMS,DBMS

Digital Equipment Corp. (DEC) offers two alternative methods for

distributed data management in the VAX/VMS environment. One method

is geared towards a decentralized environment with users having

their own processing resources and connected through a Local Area

Network. The other method is directed towards centralized

distributed processing with a VAX cluster comprising two or more

VAX processors, and the database located on a common disk. Each

of these methods can be used with either of the two DEC data

management systems: Rdb/VMS is a relational DBMS while Digital's

Database Management System (DBMS) uses a hierarchical/network

model. Further, VAXlink Client is a layered VMS software product

that provides unidirectional IBM mainframe database access and

extract capabilities; it works in conjunction with VAXlink

Answer/DB Extractor to access IMS and VSAM data structures and to

transfer extracted data to Rdb/VMS databases.

FUNCTIONAL ARCHITECTURE

VAX SQL is Digital's version of ANSI SQL and is used as an

interactive and software development interface for relational

databases that conform to Digital Standard Relational Interface

(DSRI). Used to define, access, and update relational databases,

it is characterized by an interactive data manipulation language,

2-20

Page 38: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

a data definition language utility, preprocessors and dynamic SQL

for COBOL, FORTRAN, C, and PL/I. DEC's Database Management System

uses the Network Manipulation Language (NML), a standard for

network models.

A Database Control System (DBCS) provides the connections between

the application program and the database locations for the Data

Base Management System. Application programs transmit information

to the DBCS. The DBCS transfers all information requested by a

call for data to the User Work Area (UWA). Information in the UWA

is then used to update the database.

The VAX Distributor distributes data from any Rdb/VMS, VIDA (an

access tool for IBM databases), or any other database conforming

to DSRI. The commands of the VAX Data Distributor allow the user

to specify the data to be distributed, the type of copy to be made,

the target node, and the transfer and update schedules. Three

kinds of copies are permitted by this tool: extractions,

extraction rollups, and replications. These copies effect only

the target database.

Digital's distributed, active, open-architecture dictionary is the

VAX Common Data Dictionary/Plus (CDD/Plus). VAX CDD/Plus provides

one logical dictionary system that is physically dispersed, thus

eliminating the need to develop separate dictionaries to support

proprietary tools or applications.

2-21

Page 39: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

'I1

SALIENT FEATURES

Digital claims to provide one of the best relational database

management systems for the VAX/VMS environment. VMS is a widely

used operating system for development and execution of a wide range

of applications. Apart from the above products marketed directly

by DEC, a number of other vendors have opted to select VAX hardware

as the host system for reconciling the idiosyncracies of individual

components in a distributed, heterogeneous environment.

2.10. Gupta Technologies, Inc.: SOL SYSTEM

Gupta Technologies, Inc. of Menlo Park, Calif., supports relational

database networking. The SQL SYSTEM allows applications running

on a PC to use data residing on minis, micros, and mainframes. The

primary elements of the SQL SYSTEM are: SQLBase, SQLWindows,

SQLTalk and SQLNetwork.

SQLBase is a distributed relational DBMS for PC networks. It can

be used as a standalone DBMS, as a database server for personal

computer based workstations on a network, and as a gateway to

mainframe database management systems. The SQL language syntax of

SQLBase is fully compatible with those of IBM's DB2 and SQL/DS.

Gupta Technologies claims to be the first company to offer advanced

concurrency, automatic consistency, and dynamic recovery for a

distributed relational DBMS on a personal computer.

SQLWindows and SQLTalk allow the programmer or user to utilize SQL

2-22

Page 40: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

without having to develop and compile programs. SQLWindows is an

advanced applications generator which supports graphics, menus,

dialogue and help boxes and mouse pointing devices. It contains

a complete procedural language and provides exits to other

programs. SQLTalk is an interactive data management program for

accessing SQLBase via the SQL language.

FUNCTIONAL ARCHITECTURE

The SQLNetwork components enable mini, micro and mainframe SQL

database management systems to appear like distributed database

servers on the same network. The three components comprising the

SQLNetwork are: the SQLRouter, the SQLGateway, and the SQLHost.

Requests for data are handled differently based on the location of

the databases. The SQLRouter is the software that directs SQL

requests to databases within the PC network. SQLRouter must reside

on each PC and requires NetBIOS on each workstation. The

SQLGateway routes SQL requests to databases outside of a single PC

network. It runs as a memory resident program on the PC/AT and

accesses a file containing the databases to which it can establish

a connection at load time. This file also contains the physical

node location of these databases.

SQLHost software resides on the host computer and transforms the

host DBMS into a database server that accepts SQL requests routed

to it by SQLGateway. It translates SQL commands into a format

acceptable to the host DBMS and forwards the information for

2-23

Page 41: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

III

processing. The results are then sent back to the network. A

different version of SQLHost is necessary for each different SQL

DBMS because of variations in SQL and application program

interfaces.

SALIENT FEATURES

SQLNetwork was designed to allow applications to take advantage of

the capabilities of a personal computer, to transparently access

databases that reside on diverse servers, and to serve many users

with a large amount of data. The user is provided with full

cooperative processing capabilities in DOS, OS/2 and UNIX

environments. SQLNetwork provides access to DB2 and development

is underway for SQLHost Software to support SQL/DS, Oracle,

Ingress, and Sybase.

2.11. Honeywell Federal Systems, Inc.: Distributed DatabaseTestbed System

DDTS is a testbed emphasizing modularity and flexibility for

distributed heterogeneous database management. The project began

at Computer Sciences Center (CSC) of Honeywell with the aim of

designing and building a system that could be utilized for

evaluating alternate approaches and algorithms for managing data

stored in several databases on different platforms. Although DDTS

is not a product, Honeywell utilizes this technology in specific

contracts.

2-24

Page 42: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

FUNCTIONAL ARCHITECTURE

DDTS supports a five-schema architecture. The internal schema is

divided into the global representation schema, the local

representation schema and the local internal schema. In addition,

there is an external schema and a conceptual schema which follow

the terminology of Section 1.5. A bottom-up approach is used to

design a global schema for multiple databases. First, a local

representational schema is built from the description of the local

database. Next, all local representational schemata are integrated

to form the global representation schema. Finally, the conceptual

schema describing semantic integrity constraints to be enforced by

DDTS is constructed.

The DDTS system architecture is based on a model of distributed

computing in which a set of abstract processors execute sets of

concurrent processes. Local and remote processes communicate via

message exchange. DDTS consists of a set of Application Processors

(APs) which control user interfaces and manage transactions, and

Data Processors (DPs) which manage data.

The GORDAS Command Interface (GCI) enables multiple retrieval and

update requests expressed in GORDAS, a high-level query language.

Before a GORDAS transaction is executed, the GORDAS requests are

translated to an internal representation for processing by the

Translation and Integrity Control (TIC) module. The internal

representation possesses features of relational algebra and

relational calculus, and is similar to an internal representation

2-25

Page 43: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

III

of SQL.

Materialization and access planning (MAP) result in a strategy for

processing distributed queries, updates and transactions. The

response time (the time elapsed between the start of the first

transmission and the time at which the result arrives at the

required computer) and the total time (the sum of the time of all

transmissions) are minimized with MAP.

The Distributed Execution Monitor (DEM) performs distributed

execution of a transaction as directed by MAP. Based on status

information returned by the Local Execution Monitors (LEMs) during

execution, the DEM decides to commit or abort a transaction. DDTS

uses a two-phase commit protocol during transaction execution. To

resolve concurrency control conflicts, the transaction is rolled

back and then restarted.

The Local Operations Module (LOM) of each DP acts as an interface

between the LEM and the local DBMS. LOM provides local translation

and optimization of subtransactions that have been assigned to the

DP by MAP. The LOM translates subtransaction requests in the

internal representation to data manipulation requests in the

language of the local DBMS.

SALIENT FEATURES

DDTS is a useful tool for evaluating approaches to the various

problems associated with distributed, heterogeneous database

2-26

Page 44: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

management.

2.12. IBM: DB2. SOL/DS. SAA. and others

IBM faces a challenge in developing its integrating approach. It

commands a major presence in the mainframe computer market and in

the personal computer market. It can either come up with design

approaches that enable integration of information hosted on its

own systems, or it can develop approaches that encompass non-IBM

architectures as well. So far, the emphasis has been on the former

alternative.

The IBM System Application Architecture (SAA) serves as a framework

for developing consistent applications across three IBM computing

environments: System/370, AS/400 and PS/2. SAA is a collection

of software interfaces, conventions, and protocols, and includes

support for distributed files and distributed databases that are

resident within one of these environments, and conform to certain

design guidelines.

IBM is developing a new relational Unix oriented approach that

would integrate their AIX line (System/370, RT, and PS/2) as well

as communicate with their SAA database line (System/370, AS/400,

and PS/2)). Details of this approach have not yet been released

by IBM.

At a lower level, IBM provides support for linking information from

distributed relational (SQL/DS) and hierarchical (IMS) local

2-27

Page 45: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

III

database management systems. In addition, its Data Extract product

(DXT) can extract data from DB2 tables, SQL/DS tables, DL/I data

sets, VSAM data sets, and physical sequential data sets, and

convert them for loading into other DB2 tables, physical sequential

data sets, and CMS files. DXT users can also write generic data

interfaces for accessing non-IBM (OEM) databases on MVS.

In October 1988, IBM upgraded DB2 and SQL for distributed

environments. The new DB2 and SQL (called SQL/DS) use relational

concepts to integrate information. In either case, because a

central DBMS is not assumed, each location has all the capabilities

offered by a local relational DBMS and may participate in an

overall distributed network as well. Local functions are not

affected by the role of the system in distributed processing.

Flexibility is provided for deciding how data will be distributed

and processed. The SQL data manipulation language can access

remote tables or a user may work with a local copy. Also, data may

be stored in one large table or split into several parts to

accommodate data which may be accessed and updated only at certain

local locations. Extracts, snapshots, and data replication are

various types of copies which enhance retrievals and updates.

A user may extract a copy of one database table and load it into

another database. Extracts are useful for migrating data from one

data format to another and for maintaining read-only copies of a

subset of data that rarely changes. Snapshots are read-only copies

of tables that are automatically made by the system on a periodic

basis specified by the user. These are useful for locations that

2-28

Page 46: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

need an automated process for receiving updated information on an

occasional basis. With the data replication support, any location

may update any copy of a table, and these updates are sent to other

locations automatically. This is effective for situations that

require high reliability and quick data retrieval with few updates.

FUNCTIONAL ARCHITECTURE

Distributed relational data are accessed either on a remote unit

of work or on a distributed unit of work. Remote unit of work

permits access to a single, remote relational DBMS within a unit

of work. A unit of work is a set of one or more related requests

which must all complete successfully or not at all. An application

program can query and update data from any number of tables

residing in a single DBMS within a unit of work. In addition,

application programs can access data in multiple database

management systems one at a time, in separate units of work.

SQL/DS and DB2 have remote unit-of-work capability as well as

support for single SQL requests.

Distributed unit of work provides access to multiple relational

database management systems within a single unit of work. An

application program can read and update data from any number of

database management systems within a single unit of work. However,

a single SQL statement must refer to data from a single DBMS. Key

features of distributed unit of work are: complete location

transparency, coordinated system security mechanisms to determine

whether the application is authorized to access the remote systems

2-29

Page 47: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

III

and data, and the two-phase commit protocol to ensure termination

consistency. DB2 supports distributed unit-of-work.

SALIENT FEATURES

IBM states that its leadership in stand-alone database management

products resulted from its advanced data security, data integrity,

and data recovery capabilities. Its objective is to apply this

expertise to the implementation of its distributed relational

products. DB2 and SQL/DS have referential integrity which ensures

the consistency of data values between related columns of two

different tables.

2.13. Information Builders, Inc.: FOCNET

FOCNET, derived from FOCUS Network Products, is marketed by

Information Builders, Inc., headquartered in New York City. Its

nonprocedural language and DBMS is called FOCUS. Various FOCUS

interfaces provide logical access to data in hierarchical,

relational, and network models on remote machines.

FOCNET permits multiple individuals to update a database

simultaneously. A change/verify protocol confirms that the record

has not been updated by another user before it modifies the record.

In cases where the record has been changed, the update is reapplied

against the updated version. Also, this approach allows a user to

define commit and rollback specifications for transactions.

2-30

Page 48: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

FUNCTIONAL ARCHITECTURE

A client/server model layer is inserted on top of the operating

system and the access-method architecture of each participating

node. Its three components are client software, interfaces to

network transport software and server software. This approach

reduces network traffic and takes full advantage of available CPU

resources.

Most local databases must have a Master File Description (MFD).

All files must be described in the FOCUS Master Dictionary before

FOCUS can access their data. The master dictionary provides much

more than conventional file and data information. FOCUS file

descriptions can also contain security restrictions, field aliases,

data validation criteria, predefined relational join operations,

and multiline column titles for use in reports.

FOCUS supports 4GL FOCUS, third generation languages, and ANSI SQL

global data manipulation languages. Read/write interfaces generate

the appropriate dialect of SQL for many well-known relational

database engines which use variations from this standard data

manipulation language.

Dynamic relational join operations permit linkage to up to 16

different tables and databases at one time. Processing is

optimized by including record-by-record operations and set

processing based on the functionality of the database. In

addition, balance tree indexing accelerates the speed of data

2-31

Page 49: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

III

retrieval and online updates of large databases. Users can

logically invert FOCUS databases to further optimize retrievals.

SALIENT FEATURES

FOCNET runs on many platforms under a variety of operating systems.

The list of local database management systems supported is

impressive as is the list of supported network services. FOCUS is

the most widely used 4GL in the world. It is suited for developing

software and provides for automated application development and

maintenance.

2.14. Interbase Software Corporation: INTERBASE

Interbase is a product of Interbase Software Corporation, Bedford,

Mass. Although Interbase operates in a variety of software and

hardware environments, its Digital Standard Relational Interface

(DSRI) compatibility makes it most suited for use with other DEC

DSRI-compatible tools. Relational local data models, indexed and

sequential Record Management Service files, and UNIX stream files

are supported by Interbase.

Global updates to databases can be performed in the same

interactive session, as Interbase allows multiple simultaneous

transactions per process and provides for two-phase commits. These

two features can be used to ensure that a transaction which updates

more than one database will complete correctly or not at all.

Partial commits and partial rollbacks are possible as well.

2-32

Page 50: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

Interbase uses multigenerational data structures for data storage.

Records are never updated, but only added. Old versions of the

record are deleted by the next transaction which accesses the

record. This means that update access to data being processed by

another user is allowed. On actual update, Interbase will impose

the record it originally retrieved with the record it is trying to

update. If they are different, the transaction is rolled back and

the user is informed of this fact.

Changes required to local databases are minimal. The Rdb database

of DEC permits access to Interbase files through VMS. The ORACLE

gateway requires an additional table to be defined on the database,

access to tables in the data dictionary, and two servers to be

started on the target node.

FUNCTIONAL ARCHITECTURE

Interbase supports two global data manipulation languages: SQL and

the Groton Data Manipulation Language. These two languages may be

used in precompiled third-generation language source programs or

in the Query Language Interpreter (QLI). The QLI translates either

language into BLR, a lower level language based on the DSRI, which

may be generated and read by programs.

Although InterBase has no global conceptual schema, it provides a

navigation path to distributed local database management systems

through a database handler which defines node names for multiple

databases. A program called Y-valve determines if requested data

2-33

Page 51: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

III

are on a local node or a remote node. In the case of remote data,

the request is routed through the network and the remote data

dictionary is accessed.

Interbase or RDB data requests are left in the BLR lower level

language. With ORACLE, BLR is restricted to SQL through the

gateway and passed to the ORACLE access method. Data format

incompatibility between heterogeneous systems is transparently

handled by a XDR layer.

SALIENT FEATURES

Interbase is designed to handle on-line complex applications

(OLCP). It supports: multi-user, concurrent update; automatic

transactions; serializable and high concurrency levels; two-phase

commit; multiple database transactions; and multiple parallel

transactions per process. These features provide excellent

concurrency control capability.

2,15, Massachusetts Institute of Technology: CompositeInformation Systems

The concept of Composite Information Systems (CIS) has evolved at

the Sloan School of Management of Massachusetts Institute of

Technology. The first paper on this technique was published in

1978, and significant enhancements to the original concept have

been made during the intervening years. While all other approaches

discussed in this report tend to focus exclusively on technical

issues, the CIS approach emphasizes that it is necessary to address

2-34

Page 52: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

a number of inter-related strategic, organizational and technical

issues, in order to implement systems that can effectively span

applications, functional areas, organizational boundaries, and

geographic separations.

Strategic issues include motivating cooperation between multiple

organizations, each with its own goals, priorities, and security

needs. One critical factor for such cooperation is participant

consensus on the issue of access to each others' technical and non-

technical information. Issues such as domains of shared

information, benefits to each participating group, and the role

and the responsibility of each constituent must be resolved in

advance, for the final system to succeed in a real world

environment.

Organizational issues include the process of making controlled

changes in complex organizational environments. The latest results

from the realm of inter-organizational networks have been utilized

to explain the multiple forces that modulate behavior of

individuals, groups, and organizations. Also, the notion of

focused standards has been conceived to serve as the foundation

for neutral representation as well as for the delineation of more

elaborated standards.

Technical issues relate to connectivity at both physical and

logical levels. Physical connectivity refers to the process of

actual communication among disparate systems, and covers aspects

as bandwidths, security, availability, reliability, and inter-

2-35

Page 53: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

network protocol conversions. Logical connectivity refers to the

process of accessing disparate systems in concert for composite

answers. One of the major challenges in logical connectivity is

to reconcile the different assumptions and perspectives in the

systems being integrated.

FUNCTIONAL ARCHITECTURE

The operational prototype of CIS is called CIS/TK, which is

abbreviated from CIS Tool Kit. This prototype is based on the

three-schema approach, with one integrated global schema, multiple

local schemata, and multiple application schemata. The CIS/TK

architecture includes multiple Application Query Processors (AQPs),

one Global Query Processor (GQP), and multiple Local Query

Processors (LQPs). There is one AQP for each application, and this

AQP converts an application model query, defined by an application

developer, into a sequence of global schema queries, passes them

to the GQP, and receives the results. Similarly, a Local Query

Processor (LQP) interfaces between the global data processor and

an individual query command processor (e.g., a DBMS).

The primary query processor is the GQP. It converts a global

schema query into abstract local queries, transmits them to the

appropriate LQPs, and joins the results before passing them back

to the AQP. The GQP must know how to get the data, how to map

global schema attribute names to the column names, and how to join

results from different tables. This is done with the aid of

concept agents. Conceptually, each concept agent is composed of

2-36

Page 54: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

three functional components, as follows:

(a) A concept definition which defines the rules and data,

given the global schema and an application model;

(b) A concept processor which pursues the goal of the concept

agent, given its rules and data; and

(c) A message handler which sends and receives messages on

behalf of the concept agent.

The notion of concept agents is used to perform concept

inferencing, in situations involving missing and conflicting

information.

CIS/TK utilizes an object-oriented approach. Physical

communication details and database idiosyncracies are encapsulated

within the object. The object-oriented language utilized by CIS/TK

is based on an enhanced version of Knowledge-Object Representation

Language. CIS/TK is currently providing integrated access to

several databases at MIT, and others outside MIT. External

databases include Reuter's Textline and Dataline and I.P. Sharp's

Disclosure.

SALIENT FEATURES

CIS is an integrated methodology that attempts to systematically

deal with strategic problems, organizational problems, and

technical problems that are inherent in any endeavor aimed at

integrating islands of disparate information systems. CIS/TK is

an operational prototype that utilizes innovative concepts to

perform logical connectivity in an effective manner.

2-37

Page 55: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

III

2.16. Metaphor Computer Systems, Inc.: DIS

DIS is a product of Metaphor Computer Systems, Inc. of Mountain

View, California. Through the combination of a graphical interface

and relational database technology, DIS lets users access data

from multiple databases, perform analysis, and construct their own

applications without programming. Designed for business

professionals, DIS provides tools for strategic use of data from

multiple sources; database query tools and report writers for

direct and easy access to data; spreadsheets to analyze data; plot

tools to view data graphically; and text processing tools to add

commentary to the analysis.

Currently, DIS only provides for data access to SQL relational

databases and to some flat files after they are converted to

relational format used by DIS. There is no plan to provide direct

access to non-SQL environments. However, to the extent local SQL

database management systems can translate other databases into

their own relational data model (e.g., Ingres), DIS can access

them.

There is no logical schema which encompasses all database

management systems in the network. However, a local schema of one

or several databases is graphically depicted for the user. This

provides the ability to "click" on each database and instantly see

a list of its tables. DIS can be used to query multiple physical

database management systems concurrently through multiple windows,

and to combine and manipulate such data using the spreadsheet tool.

2-3'8

Page 56: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

FUNCTIONAL ARCHITECTURE

DIS is a network of workstations (soon to include PS/2's and PC's)

and servers that interact with relational databases. Databases can

reside on Metaphor database servers as well as on non-Metaphor host

machines. Metaphor files and communication servers can provide

electronic mail and print services, plus access to a wide range of

multi-vendor environments.

The main data access tool is the Query tool. When it is selected,

the appropriate database is connected and each of its tables and

its columns are displayed in a window. As the user chooses columns

with the mouse, the Query tool draws a picture showing how the

database tables will be joined to retrieve the data. Then it

translates the graphic query into SQL and sends it to the database.

If data are not directly accessible to DIS via a gateway, they can

be downloaded to a Metaphor database server. Host resident

"formatters" are available for converting flat data files into the

relational format used by DIS.

SALIENT FEATURES

This product claims to provide a true user friendly data access.

Querying can be done without ever touching the keyboard. All

tables and columns within a physical database are graphically

provided to the user. The spreadsheet, the plotting facility, and

the word processing tools are very practical features. The list of

2-39

Page 57: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

III

government and commercial customers attests to the ability of this

product to provide management with easy access to information. IBM

has invested in Metaphor Computer Systems, Inc. and is currently

working with it to provide a similar interface for IBM mainframes.

2.17. On-Line Software International, Inc.: RAMIS InformationSystem

RAMIS Information System from On-Line Software International, Inc.,

headquartered in Fort Lee, New Jersey, is used primarily for

developing and running end-user applications. The system consists

of many diverse components such as: a centralized database, a data

dictionary, reporting and graphics facilities, links to external

database management systems, data manipulation tools, a procedural

language for application development, menu architecture development

facilities, micro-to-mainframe communications, and data transfer

facilities.

The RAMIS database may be comprised of both RAMIS and non-RAMIS

files or databases. Most non-RAMIS files and databases may be

accessed in a read-only mode. However, DB2 and SQL/DS relational

databases may be queried and updated from within the RAMIS

environment.

FUNCTIONAL ARCHITECTURE

RELATE is a facility to combine data from multiple files by

operating on them as if they were flat relational tables. RELATE

2-40

Page 58: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

provides relational capabilities regardless of the organization of

the source of the data. The user must specify the database or the

file, the data to be retrieved, the record selection criteria and

how the data are to be associated.

The RAMIS Database Manager reads and updates RAMIS, DB2, and SQL/DS

files. It allows multiple physical files to be logically

interconnected to from a single data view. A user is presented

with one relational view of the data. In the same way, multiple

physical databases may be treated as one logical database.

Specialized interfaces are required for accessing other non-RAMIS

files and databases.

RAMIS has a totally integrated, menu-based architecture front end

to all components of the system. Application developers can

utilize the menu interface, use a command language, and construct

their own customized menu interfaces.

SALIENT FEATURES

The RAMIS Information System operates on various IBM platforms

under diverse operating systems. It provides a flexible

environment for managing, retrieving, manipulating and reporting

data stored both in RAMIS format and in external databases and

files. Direct interfaces within RAMIS allow users to read non-

RAMIS files and databases using RAMIS language facilities. It is

not necessary to convert external files into a RAMIS file to access

the data.

2-41

Page 59: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

III

2.18. Oracle Corporation: ORACLE

ORACLE was developed by Oracle Corporation, Belmont, California,

to serve as a general purpose tool for providing a uniform,

integrated interface for retrieving data from several existing

database management systems. ORACLE connects with SQL/DS, DB2 and

DBase III relational databases. There is also an ORACLE utility

for 123.

ORACLE is constantly evolving to meet System R* specifications

which IBM determined in the 1970s to define the "ideal" database

management system. The influence of IBM is further seen in the

global data manipulation language which is a superset of SQL, the

IBM developed Standard Query Language. ORACLE is a widely-

installed DBMS in the VMS environment.

Both distributed query and single-node transactions are supported

by ORACLE. For updates at multiple nodes, there is no assurance

that transactions will either be committed or rolled back together.

Data loss can only be prevented by application programming

techniques. Global updates and the two-phase commit procedure will

be supported in the next product release.

FUNCTIONAL ARCHITECTURE

The Open Systems Architecture of ORACLE for distributed processing

and distributed database management is SQL*Star. It is comprised

of three ORACLE products: SQL*Net, Distributed ORACLE relational

2-42

Page 60: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

DBMS, and SQL*Connect.

SqL*Net provides distributed processing capability, and the

application software can run on a different machine from the ORACLE

kernel. Users can interface SQL*Net to arbitrary network

environments. SQL*Net is communications protocol independent. It

consists of a generic layer which is common to all protocols and

environments, and a custom layer which is specific to each protocol

and operating system.

The Distributed ORACLE relational DBMS supports location

transparency, single site transactions, multi-site updates in

multiple transactions and site autonomy. To meet System R*

specifications, ORACLE is planning multi-site updates in a single

transaction, replication and table partitioning for future

releases. Distributed data dictionaries contain information about

local data and how to get access to global data.

SQL*Connect allows ORACLE to communicate with non-ORACLE database

management systems. Plans for broadening access to other database

management systems, as well as to non-database file systems (VSAM,

Record Management Service) are continuing. Query optimization is

achieved by using an array interface to transfer data in batches.

This insures that the network is used efficiently.

SALIENT FEATURES

ORACLE products run on an impressive range of mainframes, minis,

2-43

Page 61: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

III

workstations and personal computers (over 90 hardware platforms),

and support a large number of operating systems and a wide variety

of network services. The designers intend to encrypt usernames,

passwords, and SQL data. Approaches for meeting B1 and C2 National

Computer Security Center (NCSC) security requirements are under

development.

2.19. Relational Technology. Inc.: INGRES

INGRES is marketed by Relational Technology, Inc. of Alameda,

California. Products of this company are used extensively around

the world. Approaches originally designed for homogeneous

environments have been extended to cater to heterogeneous

situations.

FUNCTIONAL ARCHITECTURE

The INGRES approach consists of: INGRES/NET Network Protocol

Support, INGRES/STAR Distributed Data Manager, and INGRES Database

Gateways. INGRES/NET is a transparent software layer that provides

INGRES protocol support for DECnet, TCP/IP, SNA and ASYNC

communications. This software allows access to databases across

a network and performs automatic data conversion. With the

INGRES/STAR Distributed Data Manager, simultaneous access to

distributed databases across a network is possible. INGRES/STAR

provides complete location transparency and local autonomy. INGRES

Database Gateways enable other types of databases and file systems

to be included in a distributed database. The appropriate gateway

2-44

Page 62: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

module must reside on the local host. Relational, hierarchical,

and network local data models are supported.

A two-phase commit protocol is used for distributed transaction

processing. If a site involved in a transaction is inaccessible,

other sites are automatically rolled back. When a system crashes,

INGRES backs out all pending transactions and returns the database

to a consistent state without operator intervention.

The global data manipulation language of INGRES is ANSI SQL. After

receiving a SQL query from an application, INGRES/STAR breaks it

into subqueries and directs them to the appropriate local data

managers and gateways. Next, each local data manager executes its

respective subquery and the data are transferred to INGRES. Data

can be transparently joined and combined with INGRES SQL end-user

interfaces or application development tools.

The INGRES/STAR Distributed Data Manager combines distributed query

optimization with parallel processing. The distributed optimizer

selects the best route for satisfying a multicomputer request.

This approach reduces network traffic and improves performance.

With this open, multi-server architecture, system administrators

can customize their environment by assigning various servers with

dissimilar priorities for different tasks, and by connecting

transaction applications to a high priority server and batch

report-oriented applications to a lower-priority server.

INGRES allows dynamic control of parameters to adjust the

2-45

Page 63: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

III

concurrency of the system to the needs of the application set.

Transaction and decision support applications may be executed at

the same time. Also, the multi-server architecture allows system-

wide monitoring and control of database sessions for real-time

system management. A virtually infinite number of simultaneous

users is possible.

The INGRES data dictionary stores all performance related

information-table structures and sizes, available indexes, and data

distribution statistics. The more information is provided in the

dictionary, the more intelligent is the Query Optimizer. The Query

Optimizer produces query execution plans which can be examined to

determine the effects of adding indexes, selecting different

storage structures, and adding data distribution statistics.

SALIENT FEATURES

INGRES incorporates many of the features required for distributed

transaction applications including two-phase commit, automatic

recovery, and table fragmentation. Distributed query optimization

strategies and full parallel query execution capabilities enhance

its distributed on-line transaction processing capability.

2.20. Software AG: ADABAS

The ADABAS DBMS, developed by Software AG of West Germany, is a

part of an open Integrated Software Architecture (ISA). Used in

conjunction with NET-WORK (for distributed processing) and ADA-

2-46

Page 64: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

NET (for distributed databases), ADABAS can support heterogeneous

networks involving IBM, SIEMENS, DEC-VAX and WANG computers. The

ADABAS DBMS supports relational, hierarchical, and network local

data models.

ADABAS is designed for production, development, and information

center environments. It claims to handle both high transaction

volumes and complex ad hoc queries operating against large volumes

of data. Distributed transactions can be in the form of a series

of updates to different databases which might involve more than one

network node.

FUNCTIONAL ARCHITECTURE

ADABAS uses either SQL embedded in COBOL, FORTRAN or PLI or NATURAL

as its global data manipulation language. NATURAL is a fourth

generation language based on a Communication Interface and standard

protocols. NATURAL applications may be ported without modification

across a wide range of system environments and hardware

architectures.

The ADANET product of Software AG allows a user level logical

ADABAS call to be translated into one or more physical calls in a

transparent manner. A dynamically maintained ADANET translation

table defines logical databases and their corresponding physical

database representations. Three logical file structures are

supported by ADANET: replicated, distributed and isolated files.

2-47

Page 65: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

III

PREDICT is the centralized integrated active data dictionary. The

structure of the metadata in PREDICT describes the logical and

physical interconnections of all data and programs to be managed.

PREDICT offers a user-friendly, menu-driven user interface for the

management of data and application information.

SALIENT FEATURES

The Open Integrated Software Architecture (ISA) provides a high

level of functionality in areas such as end user access,

application solutions, application development and database

management. It also provides software independence across multiple

hardware and operating platforms without requiring any change in

local databases. Global update processing is supported for

relational, network and hierarchical local data models. The global

data manipulation language of ADABAS reduces dependencies on third

generation languages and is highly portable.

2.21. Sybase. _Inc.: SYBASE

SYBASE is marketed by Sybase, Inc. of Emeryville, California. It

can handle data in either hierarchical form or relational form.

Data in a network DBMS are accessed via a relational interface.

FUNCTIONAL ARCHITECTURE

The Sybase Distributed System is comprised of three major

components: the Sybase Data Server, the Sybase OpenServer, and

2-48

Page 66: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

the Sybase StarServer. The Sybase DataServer allows remote queries

and distributed transactions across multiple DataServers while the

Sybase OpenServer extends this capability to other applications or

DBMS Servers, including non-SQL systems. The Star Server adds the

functionality of site transparency, including cross-site joins and

views, and the Distributed Data Dictionary. The SYBASE OpenServer

is in the final stage of beta testing; the DataServer and the

StarServer are commercially available.

The SYBASE DataServer allows multi-site update transactions, while

ensuring the logical consistency of the data via a two-phase commit

protocol. A 'commit service' running at a designated site logs the

progress of a transaction. The commit service logs notes about

when all sites will be prepared. If the transaction fails before

that point, it will be backed out; if it fails after that point,

it will be rolled forward.

With its multithreaded client/server architecture, SYBASE assumes

many of the functions of an operating system. SYBASE avoids the

overhead of process swapping by scheduling and activating users.

Multiple users share a single copy of the SYBASE image, data and

procedure cache, and transaction log. Because each additional user

requires only 30 KBytes of memory, hundreds of users can

simultaneously access and update data on any node in a distributed

network, according to Sybase.

A Virtual Server Architecture (VSA) for Symmetric Multiprocessor

Systems is scheduled for beta testing. The VSA treats processors

2-49

Page 67: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

III

as resources to be allocated on an as-needed basis. An application

or a user is not restricted to a specific processor; instead

applications may run on various processors at different points in

time depending on the number of blocks for I/O requests or other

reasons. Four key features of VSA are: synchronization control,

automatic load balancing, optimal resource utilization, and dynamic

configuration.

SYBASE uses a Distributed Data Dictionary, consisting of synonyms

which are used to locate objects, and Remote Object Catalogs which

contain information about the characteristics of those objects.

With this approach, users and applications need not be concerned

with changes in the physical location of an object.

SALIENT FEATURES

Sybase is a leader in secure database management systems. The

SYBASE Secure SQL Server meets Division B security requirements

defined by the National Computer Security Center. In compliance

with Division B specifications, SYBASE offers the ability to store

data of multiple security classifications in a single database.

Because networking was built as a key part of SYBASE, rather than

as an add-on, network usage is optimized. The network is used more

efficiently by sending a single 'execute procedure' command,

instead of an individual SQL statement.

SYBASE allows server-to-server communications. Multiple users may

2-50

Page 68: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

share the communication link between servers, thus eliminating the

overhead of opening and closing a linkage for each user.

2.22. Tandem Computers Incorporated: NonStop SOL

Tandem Computers Incorporated, headquartered in Cupertino,

California, offers software and network compatible systems ranging

from mainframes to low-cost network nodes. Because all Tandem

processors use the same operating system and distributed database,

as many as 255 systems with a maximum of 4,080 processors can be

linked to form a geographically dispersed network. An individual

system can support hundreds of terminals and a database with

billions of bytes of data.

FUNCTIONAL ARCHITECTURE

GUARDIAN 90, a message based operating system, supports the

parallel architecture of Tandem NonStop systems. Key features of

GUARDIAN 90 are: fault tolerance, distributed data processing

capabilities, high transaction throughput, linear expandability,

security, data integrity, and the ability to link systems,

resources, and devices.

GUARDIAN 90 supports fault tolerant process pairs. In the case

where a system software error causes a process to fail, its backup

process takes over in another processor. In addition, GUARDIAN 90

manages communication between processors. Database availability

is always ensured. Mirrored volumes of data are stored on a pair

2-51

Page 69: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

III

of physically independent disk drives. When a failed drive is

operational again, the system automatically brings it up to date.

The systems resume normal mirrored operation after the restored

disk is fully updated. Via its message system, GUARDIAN 90

coordinates all distributed activity in a fashion that makes all

data accesses transparent to the user. An application can read,

write, and update data regardless of the physical location of the

data. A built-in password security system is used to control file,

device and process availability. Tandem offers integrated security

and encryption products to protect information from unauthorized

access.

SALIENT FEATURES

The multiple parallel processing capability, the fault tolerance

capability, and the expandability feature make NonStop SQL a good

product for on-line transaction processing (OLTP).

Tandem Corp. has signed integration agreements with Oracle Corp,

Sybase, Inc., Relational Technology, Inc. (Ingres), and Information

Builders, Inc. (FOCUS) to port their front-end tools to Tandem

Systems. Details of the approaches used by these companies are

presented in their respective subsections.

2.23. TRW: TDIE

TDIE is a data integration engine developed by TRW, Redondo Beach,

California. Unlike conventional efforts which use a gateway with

2-52

Page 70: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

query language mapping, TDIE has no global data manipulation

language. A local DBMS is used and a connection to a remote DBMS

is triggered when a user issues a demand requiring data

integration.

TDIE is not oriented towards any particular environment. No change

is required in existing applications. Additional local database

management systems may be added to the system configuration and

more than one global schema is permitted.

An integrated scheme performs uniform query and updates for

hierarchical, relational and network data models. Automatic

updating is performed by polling the potential source files at a

rate suitable for a given data category. Types of data that are

frequently accessed and must be current are polled at short

intervals. This approach eliminates unnecessary activity and

avoids unnecessary copying of bulk data by transferring only the

changes in data.

FUNCTIONAL ARCHITECTURE

The TDIE is comprised of the Data Transform Manager and the Host

Interface Manager. These may reside on one processor or they may

be partitioned across multiple processors.

The Data Transform Manager contains the centralized dictionary

which defines the global conceptual schema. This knowledge-based

dictionary supplies scripts, translation software, and data

2-53

Page 71: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

III

locations for full data integration. Scripts are instructions for

data transfer, which are sent to the Host Interface Manager, while

the translation software contains the translation of one query

language to another.

To add another local DBMS, translators and scripts must be added

to the dictionary. TRW provides a front end tool called the

Integration Advisor for interactive construction of a global

integrating schema.

The Host Interface Manager is the bridge to the host applications.

It provides for automatic updating and data retrieval. Time-

critical applications in transaction processing or real-time

control are not supported. Also, because TDIE is not a distributed

heterogeneous DBMS, concurrency control is an important issue.

SALIENT FEATURES

TDIE supports global updates and enables the integration of

hierarchical, relational, and network data models. It is layered

on top of the existing systems and does not require any change to

them. New systems can be added as requirements change. Also, the

TDIE can be implemented in increments as desired.

2,24. Xerox Advanced Information Technology: Multibase

Multibase was developed by Research and Systems Division of

Computer Corporation of America. This division was acquired by

2-54

Page 72: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

Xerox and is now called Xerox Advanced Information Technology

(Xerox AIT). Although Multibase is a prototype and not a product,

Xerox AIT utilizes this technology for providing customized

distributed heterogeneous database management systems for its

clients.

Multibase provides a uniform, integrated interface for retrieving

data from several existing database management systems. It allows

a user to reference data in heterogeneous databases, through a

common query language, using a single conceptual schema. Multibase

is designed to serve as a general tool, without specific

orientation towards any particular application area. It allows

existing applications to operate without change and also permits

new local systems to be included in an existing multibase system

configuration.

The integrated access available through Multibase does not provide

either the capability to update data in local databases or the

ability to synchronize read operations across several sites. In

order to process user queries, the system must request and control

specific services offered by the local systems (e.g., locking local

items).

FUNCTIONAL ARCHITECTURE

Multibase uses the language DAPLEX as its global data manipulation

language. DAPLEX provides constructs that allow users to model

real world situations in an efficient manner. A three level schema

2-55

Page 73: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

III

of definitions is employed as discussed in Section 1.5. The

process of global information retrieval involves two main

components. These are: Global Data Manager (GDM), and Local

Database Interface (LDI). The GDM performs tasks of Command

Processor, Decomposer, Merger and Distributed Execution Monitor.

The LDI design of Multibase is based on the needs of a local DBMS,

for example, a more sophisticated LDI is designed to support a file

system as compared to the one that supports a DBMS.

SALIENT FEATURES

Multibase provides an integrated scheme for uniform query access

to dissimilar systems including hierarchical, relational, and

network local data models, and File Systems. Global query

optimization is performed in Multibase.

2-56

Page 74: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

SECTION THREE: DISCUSSION AND CONCLUSION

The first section of this report explained the basic principles of

integrating disparate information systems. In the next section,

a broad spectrum of commercially-marketed products and approaches

were discussed. Based on the material covered so far, this section

develops a set of recommendations for selecting the optimal product

or products. In particular, the information presented in Section

2 and Appendix B is utilized to develop a taxonomy for classifying

all the alternative approaches.

In Section 1.4, it was emphasized that changes to existing

information systems should be as little as possible; in the ideal

case, no change should be required. Subsequently, in Section 1.5,

the importance of a single global conceptual schema was emphasized.

Many products and approaches tabulated in Appendix B indicate that

multiple global schemata can be created to suit the needs of

various users. From a technical viewpoint, these are external

schemata, not global schemata. These are some of the

considerations that were examined while developing the broad

categories described in the succeeding subsections.

3.1. Unintrusive Integration

This is the most elegant form of integrating information systems.

The word "unintrusive" connotes that the integration methodology

does not demand that specific modules be resident in each

3-1

Page 75: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

III

participating information system (see Figure 1). There is a

definite assumption in products belonging to this category that

existing information systems cannot, and should not, be tampered

with. On top of these systems, a single global conceptual schema

is introduced.

Use of this approach in a multiorganization or a multidivision

environment implies that the existing information systems of

various divisions and organizations will not require any

modification. These systems will be accessed by a central system

that holds a global schema with names and locations of all data

items in participating systems. A user will access the central

system, which will retrieve information from appropriate sources.

The use of a central system and a single global schema offers

appropriate controls on data integrity, access rights, and the

feasibility of developing a new information system over time.

The main disadvantage of this approach is that is difficult to make

global updates to multiple systems as there is no control over the

capabilities of individual information systems. The application

program must, therefore, be designed to safeguard the integrity of

the updates.

Notable products in this category are: Multibase (developed by

Computer Corporation of America and now marketed by Xerox);

Mermaid (developed by Unisys and now marketed by Data Integration,

Inc.); and CDM*Plus (developed under the IISS initiative of U.S.

3-2

Page 76: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

4=

/A L

0U-

a-

EU)U)I-c:

2o00

0

00

%"0

Cu

a)C:oc0 c

C0

I

0)

a

a)Im4-anMI

0-

L,m_z

AL

/

�I

=3�=31

1,

4wI

G-m

Page 77: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

Air Force and now marketed by Control Data Corporation). The

development of all these three products involved huge sums of

money, and the acceptance of these products is still not

commensurate with the investment. Defense organizations have been

in the forefront in evaluating and field-testing these products.

3.2. Intrusive Integration

Approaches in this category require some modification of existing

information systems or the location of specific modules within

these systems. These modules are typically utilized to perform

network services, to store information on location of specific

files, or to initiate secondary processes in case some information

resources are not operational or unusually slow in responding to

requests (see Figure 2). If these modules are designed and

implemented properly, and if the existing information systems offer

the flexibility to accept these modules from technical and

organizational considerations, then this approach can offer

significant performance improvements over the approach outlined in

Section 3.1. The main drawback, however, lies in the fact that

existing information systems frequently use older technologies,

employ outdated programming practices, offer little good

documentation, and can only be modified after prolonged effort.

Notable products in this category include SUPRA (original

technology developed in West Germany and now marketed by CINCOM),

INGRES (from Relational Technology Inc.), and Sybase (from Sybase).

3-3

Page 78: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

L

/e

k

A

U,

E

Co0

Em.

Cu00-

n00

a)c)C

en

O - E0 0 n

aU)

C0

CuI..

a

C)I.

Nom

CD

0I:

IL

A

/

- ^~~·

1_1

it--

I

qfiI

CZ=-

Page 79: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

III

In the case of Sybase, it appears that the present family of

products support SQL databases only, and support for non-SQL based

databases and file systems is still being tested. There are some

restrictions on the types of data models and data formats supported

by these three products. For example, Supra requires data to be

in first normalized form and keys to be in ascending or descending

order. CINCOM is interested in developing a common protocol for

integrating heterogeneous databases, that could be utilized by

multiple companies. Because of these reasons, it is important to

analyze potential applications carefully to determine which of

these products, if any, is appropriate in the short-term and in the

long-term.

3.3. Interfacing with Controlled Redundancy

In this approach, no effort is made to create a global conceptual

schema. Instead, based on the perceived needs of the user,

strategies are designed to get the necessary pieces of information

on a "just-in-time" basis (see Figure 3). As an example, consider

a hypothetical application which requires that information on

100,000 items (which could be key customers, spare parts, or

inventory levels) be available every morning. Assume that the

existing information systems that are used to track the position

of these items are updated with different periodicity: very few

on a yearly basis, some on a monthly basis, a large percentage on

a weekly basis, another large percentage on a daily basis, and the

rest on a six-hourly basis. In such a situation, information can

3-4

Page 80: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

CI

13'0

G)or.

a)

.r40c .

I

C:

mLo0

'-*0C

C

C)

C)

Page 81: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

III

be obtained in advance; further, to fulfil anticipated requests

one can get information only for the items that have changed since

the previous status report.

One product that exemplifies this approach is Data Integration

Engine from TRW. In this patented approach, when a change in the

status of a source information system occurs, this change is

transmitted to the destination application with a priority that

would allow it to reach the destination before the next scheduled

time of retrieval. The software to accomplish this task and other

related tasks is comprised of two types of material: a basic data

engine, which is site independent, and additional material which

must be customized to each site. The basic data engine is written

in C and can operate in a UNIX or a VAX environment. The

additional material can be generated with an expert system called

Integration Advisor, which was originally written in LISP and has

recently been converted to C. According to a technical

representative of TRW, the ratio of the size of the basic data

engine to the size of the additional material is 5:1.

Incidentally, TRW has recently been awarded a multiyear contract

by the U.S. Department of Defense.

3.4. Interfacing with Transaction Transfer

This is the simplest technique for managing transactions in a

heterogeneous information system environment. Instead of creating

a global schema or prestaging data to meet expected needs, this

3-5

Page 82: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

approach implies that action begins after a user enters a command.

Unless the user is willing to tolerate long delays, this approach

is good only in situations where only a small number of information

systems need to be accessed in response to the user query (see

Figure 4).

Among the notable products in this category, the one that deserves

mention here is DIS from Metaphor Computer Systems, Inc. DIS

possesses two unique characteristics. First, it is completely

icon-driven and as such it is very easy to learn and use. Second,

IBM has invested in Metaphor Computer Systems Inc. and they are

working together to provide a similar interface for IBM mainframes.

As such, the DIS interface has the potential of becoming a commonly

used standard for dealing with multiple information systems.

3.5. Additional Evaluation Criteria

In addition to determining which of the four distinct approaches

described in Sections 3.1-3.4 above may be most appropriate for a

particular application, it is necessary to examine a number of

other technical and non-technical issues before embarking on a

specific implementation strategy. These issues are delineated in

the following paragraphs.

3.5.1. Composition of Transactions

Transactions are of two types: queries and updates. In our

3-6

Page 83: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

a)I-C,

oU)

I -

CI cM

4lNor.

0U.

cm:

ZL

III

AL A

-- -T

L

Ir

Page 84: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

context, a query involves reading of data in multiple computer-

based systems, manipulating the data retrieved to make them

consistent, and then presenting the result to the user. The

original data in source computer-based systems remain unaltered by

a query. An update, however, results in modification of data in

various information systems. It is much more difficult to deal

with updates than with queries. If an application involves

updates, then a significant number of the alternative approaches

discussed in Section 2 should be deleted from consideration. The

ones to be deleted can be readily identified from Appendix B, which

contains a specific question on how update transactions are

managed.

3.5.2. Security Considerations

When one makes it easier to automatically access relevant pieces

of information in disparate systems, the same capabilities become

available to any unauthorized user of the system. Any integrated

set of systems is more vulnerable to security violations than a

single information system.

None of the vendors was specifically asked to comment on this

issue, although a few of them have provided information on their

compliance level with security requirements defined by the National

Computer Security Center. This issue may hold significant

weightage in the case of large defense, banking, and insurance

applications.

3-7

Page 85: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

III

3.5.3. Corporate Expertise

Most large computer corporations have tended to concentrate on

integrating systems of their respective makes only. For example,

in Section 2, there was a discussion of IBM and DEC suggesting

solutions pertaining to their respective families of products.

Frequently, smaller companies come up with non-traditional

concepts, and larger companies tend to endorse them after these

concepts have attained some degree of maturity and acceptance by

the market.

The fact that the companies who developed good prototypes such as

Multibase and Mermaid later spun off their development groups must

be viewed with caution. This is an evolving field, and user

organizations must be careful about selecting only the set of

products and services that are likely to be supported by more than

one company.

3.5.4. Applications and Development Strategy

In the current business environment of increasing globalization

and corporate reorganizations, applications and functions are apt

to change in relatively short periods of time. The present

technical report has been written without any specific application

in view. As an organization determines which applications are

important in the short-term and the long-term, it would become

feasible to make more specific comments about the relevance of

3-8

Page 86: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

different products and approaches.

It is important to note that any endeavor towards integration has

ramifications for future integration initiatives. The selected

approach must offer the flexibility to be enhanced in a modular

fashion. Further, public-domain approaches should be preferred

over proprietary ones. Based on these considerations, a

development strategy can be delineated.

3.6. Conclusion

The integration of existing information systems involves a careful

analysis of many technical, organizational, and strategic issues,

and the surmounting of many technical and nontechnical barriers.

This technical report has focused on products and their

capabilities related to logical access to relevant pieces of

information. For discussion of other approaches and issues, the

reader may wish to refer to Ref. [2] and the set of eight books on

Knowledge Based Integrated Information Systems Engineering

published by M.I.T.

The problem of heterogeneity can be overcome either by integration

or by interfacing. To decide which option to select, it is

worthwhile to look back to the stage when the concept of database

management system originated. If one wanted to deal with a single

clearly-defined application, it was better to write a customized

program to do so; however, if there were many users with diverse

3-9

Page 87: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

III

needs, it became increasingly attractive to use the concept of a

database management system, even though it involved a larger

investment at the initial stage. The same considerations apply at

the metalevel today. If access to multiple information systems is

needed exclusively for a small set of clearly-defined applications,

the interfacing alternative should suffice. However, if there is

likely to be an array of broad applications, it appears preferable

to take the integration alternative. Remember that "just as the

evolution of integrated database management technology displaced

conventional piecemeal programming approaches in the 1960s,

similarly the new information integration approaches will lead to

a significantly different computational environment by the year

2000 A.D.!" [1].

3-10

Page 88: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

REFERENCES

[1] Gupta, Amar, Integration of Information Systems: BridgingHeterogeneous Databases, IEEE Press, New York, 1989.

[2] Morris, J.H., Satyanarayanan, M., Conner, M.H., Howard,J.H., Rosental, D.S.H., and Smith, F.D., "Andrew: ADistributed Personal Computer Environment," Commun. ACM,Vol. 29, 3, March 1986, pp. 184-201.

[3] Balkovich, E., Lerman, S., and Parmelo, R.P., "Computingin Higher Education: The Athena Experience," Commun.ACM, Vol. 28, 11, Nov. 1985, pp. 209-215.

[4] Adiba, Michel, and Portal, Dominique, "A CooperationSystem for Heterogeneous Data Base Management Systems,"Inform. Systems, Vol. 3, 1978, pp. 209-215.

[5] McLeod, Dennis, and Heimbinger, Dennis, "A FederatedArchitecture for Database Systems, " in AFIPS Conf. Proc.,Vol. 49 (National Computer Conf., 1980), pp. 283-289.

(6] Lam, Chat Y. and Madnick, Stuart E., CompositeInformation Systems - A New Concept in Information System,Working Paper #35, Center for Information Systems Research,Sloan School of Management, Massachusetts Institute ofTechnology, May 1978.

[7] Gupta, Amar, and Madnick, Stuart, Knowledge-Based IntegratedInformation Systems Engineering: Highlights andBibliography, Massachusetts Institute of Technology, 1987.(NTIS/DTIC Accession # A 195850).

[8] Litwin, Witold et al, "SIRIUS Systems for Distributed DataManagement," in Distributed Databases, edited by: H.Schneider, North-Holland Publishing Co., 1982, pp. 311-366.

[9] Heimbinger, Dennis, and McLeod, Dennis, "A FederatedArchitecture for Information Management," ACM Transactionson Office Information Systems, Vol.3, No. 3, July 1985, pp.253-278.

[10] Thompson, Glenn Ray, Multi-database Concurrency Control,Doctoral Dissertation, Oklahoma State University,Stillwater, Oklahoma, 1987.

[11] Madnick, Stuart E., and Wang, Y. Richard, "IntegratingDisparate Databases for Composite Answers," in Proc. ofthe Twenty-first Annual Hawaii International Conferenceon System Sciences, Vol.II, January 1988, pp. 583-592.

[12] Pu, Carlton, Superdatabases for Composition of HeterogeneousDatabases, Technical Paper No. CUCS-243-86, Columbia

Page 89: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

1'

University, New York, 1986.

[13] Gibbons, Phillip B., "A Stub Generator Multilanguage RPC inHeterogeneous Environments, IEEE Transactions on SoftwareEngineering, Vol. SE-13, No. 1, January 1987, pp. 77-87.

[14] The ANSI/XI SPARC DBMS Framework Report of the Study onDatabase Management Systems, AFIPS. 1977.

[15] Appleton, Daniel S., "The Technology of Data Integration,"Datamation, November 1, 1985, pp. 106-116.

[16] Bhalla, S., Prasad, B.E., Gupta, A., and Madnick, S.E., "ATechnical Comparison of Distributed Heterogeneous DatabaseManagement systems," in Integrating Distributed Homogeneousand Heterogeneous Databases: Prototypes, edited by AmarGupta and Stuart Madnick, M.I.T., Cambidge, MA, 1987, pp.161-218.

[17] Reddy, M.P., Prasad, B.E., and Reddy, P.G., "QueryProcessing In Heterogeneous Distributed Database ManagementSystems" in Reference [1] above, pp. 264-277.

[18] Thompson, Paul, "Reference Models for InformationEngineering," in Reference [1] above, pp. 291-291.

R-2

Page 90: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

BIBLIOGRAPHY

[1] Gupta, Amar, and Madnick, Stuart, eds., KBIISE--Vol. 2:Knowledge-Based Integrated Information SystemsDevelopment Methodologies Plan, M.I.T., Cambridge, MA,1987 (NTIS/DTIC Accession # A 195851).

[2] Gupta, Amar, and Madnick, Stuart, eds., KBIISE--Vol. 3:Integrating Distributed Homogeneous and HeterogeneousDatabases: Prototypes, M.I.T., Cambridge, MA, 1987(NTIS/DTIC Accession # A 195853).

[3] Gupta, Amar, and Madnick, Stuart, eds., KBIISE--Vol. 4:Object-Oriented Approach to Integrating Database Semantics,M.I.T., Cambridge, MA, 1987 (NTIS/DTIC Accession # A195854).

[4] Gupta, Amar, and Madnick, Stuart, eds., KBIISE--Vol. 5:Integrating Images, Applications and CommunicationsNetworks, M.I.T., Cambridge, MA, 1987 (NTIS/DTIC Accession #A 195854).

[5] Gupta, Amar, and Madnick, Stuart, eds., KBIISE--Vol. 6:Strategic, Organizational, and Standardization Aspects ofIntegrated Informations Systems, M.I.T., Cambridge, MA, 1987(NTIS/DTIC Accession # A 195855).

(6] Gupta, Amar, and Madnick, Stuart, eds., KBIISE Vol. 7:Integrating Information Stystems in a Major DecentralizedInternational Organization, M.I.T., Cambridge, MA, 1987(NTIS/DTIC Accession # A 195856).

[7] Gupta, Amar, and Madnick, Stuart, eds., KBIISE--Vol. 8:Technical Opinions Regarding Knowledge-Based IntegratedInformation Systems Engineering, M.I.T., Cambridge, MA,1987 (NTIS/DTIC Accession # A 195857).

[8] Wang, Richard, and Madnick, Stuart, eds., ConnectivityAmong Information Systems, CIS Project, M.I.T., 1988.

[9] Special Issue on Distributed Heterogeneous Systems,Proceedings of the IEEE, Volume 75, No. 5, May, 1987.

[10] Bachman, Charles W., and Ross, Ronald G., "Toward a MoreComplete Reference Model of Computer-Based InformationSystems," Computers and Standards, Vol. 1, North-HollandPublishing Company, 1982, pp. 35-48.

[11] Cole, Robert, "A Method for Interconnecting HeterogeneousComputer Networks," Software-Practice and Experience, Vol.17, No. 6, June 1987, pp. 387-397.

[12] Gavish, Bezalel, and Pirkul, Hason, "Computer and Database

B-1

Page 91: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

'II

Location in Distributed Computer Systems," IEEE Transactionson Computers, Vol. C-35, No. 7, July 1986, pp. 583-590.

B-2

Page 92: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

Appendix A

COMPANIES CONTACTED AND APPLICABILITY

Massachusetts Institute of Technology, 1989. All rightsreserved. Reproduction of tables or quotation form themprohibited. All requests for grant of permission should bedirected to the principal investigator, Dr. Amar Gupta.Telephone (617) 253-8906.

I

0

Page 93: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

III

C. I0 C

,)I 0

o0

lC aoo

0

4(U 04.

>0 (

- L 44144 E-4 0

.,4

a

0

C 0-c

o 10

4-) MI a

aI

o .U o 0 o

0.,I E -

J kC * ~OaUl = > _P H; DCC

a n C

oa.CU

I0*-H 0a4J4 04 4C c

0 to .4

Z

n

I

'-4OMt

> 0

a,

mo 0

_

04) 0la C

> >

I >1

CO C..) : Ca

x C.4 a 94

-4 0 t 0 C

,4 .,

W -H 0I.(U·4) a

0 4- 04 3 O a

' O C t

0 4-,'4 ( ~.Z '~ -)o

C,1

a:

Q. ~ IO U O4O

N C

.*-4 > o

04-4-

0 00 0

4 -' C

C.0

341- o

I

0.-4

a,

C

* 4

40C C

4 *U43

C9 j

4-IN Cq

.4 ., 0 Ca* O1 W I

a

m co rC so C ) co c o

4. 0 C O C 40 U, a

4 O 4 >

> 0C H

'U ao 04

C O 0

0

C 0

Z ) 0 Sq > 0 O

m IEQ v

>4E-4H04

H

H1.4Of

E4= w

a E4O CnCs: >P4 to

>4

He(

C.)H

zzXP <E4

zCnloHz W E-4

0UIcn ~wi

1241:0C.

rn

0Zr20 I la

E-4U

0iz0U

>4z04

0C)

: --� i -

Page 94: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

--- - - ,,, ,,,

4n>.~C

waH

aU

,

~0 3·

~.~mPa O3 C

Sr al- U:It.,{ C HI u4i P-4 -P C IC _4 = o W(

0>- C 4JM J 4) en Q)H a: o QcH o

Q) 0 L·0o Ui c ow..q., or ·r: C:a u: ff sp :D p

P:cQ o

fet *-1

E

8 = cO C

ma

aPIc

4)C;

4i0

0M

qTJ :-

e..~ ~ ~ .I

aRv C

u C0

0 C

i Pc

oao

_

ar:U

0M.4

oo1

co r,raO

vl~~~~l

e) )

M 0

eIIa -4~Dor- I I

0b( Qo Q oU E- cD 0

I· ,-i VE E tUa a

C a- ....

3 nP CQor E O Fr

tn N v ffi

· r O

Ss C aE x a V OU X; (

MaCO

I

C -1X

p a: o o01 O

v C I Iv H 0 o

C : 03 CO

aca cto c-W & O q D

M Qq " 4 L rp4 0 Si C (L X

Q tS 0 y

c*,

3

I

lc

0:C2UC

EH- c· vo

C: ,'1erl

.Io *-a a-44) rM )

I

10

cC:.el

M-

·

aC Co9

c CC c~'"0a i>C::

P,'a0) "a

I0

* "4r.r2

-.-

o

·C0C

* V

4. 0

C P

-

a a

cJI L0

tCEn4rOJ3 3

C .rq

403., - j*X 4M MC 4

O) .

C 4) :3

>4

>4E-4Hc

14

HMr

Pc

>4

H

go

UCH4

x CaH1CIm IWI ww E-4(04 ri

04 4 -

z0nUMWHcz4

u

U a

M

tz n

a

w

u G 4

z

Pc

Z04

Prz:

u

Page 95: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

111

I

C)z

.-4u

0C

.H.r(

(a4)

C

(14)'Alw

4JC4)2

00

ra

C.9-

4)

0

4-))

0H r'OCN

-0 0n I Q0 o2

0 0>iO Lmm; H o 0 4n

r a 0 .0 C

U) c4 a

0 0

m I C0 0

0 ~4-H4.) >

(004) C Cot

.4 w

0 m

04) 4)ZOq IN

4-)

LOO

00 X

.0 - I I~ .~ 0 ,.D .·) C ,

r

m 0

M U_H 0I

e -H =

D10

L.4 4. 4.)

4) CH C

Q 0 4)C)- -

O

W O o )

c > C a4)4)0

- 0).

r

Q m 4mOrO X

*- 0 5;

· -4.4X O

4)cnM

0l

C4-)

4)a

Q

>.9-44) 00

FO ) P HO 5-- 4 f0Ct

0 4 r")f20 4)

40) CiW0 DW

(L 4 O --i=

0 It·cv94 -i r .

;4)m03 00rz 4) O

0 0 e4tu O U

C 4)e-40X O

UO <.9-4 05-4 _. .H_cr Ue ne

Srd

00 -U o

>1 4)

13

O CJ OOC 0 44002

4.r

4)0

·4. .r.O ~

4)

O0X L

0 0

L r)> 4)

0 C

0 0 -H.4.-H -)

i( NE q

> C -4I o-, ',OQ m.-. O

I IC

De 1-4 C:, r. r,

'~ 3 1~ o~

O 0 O

$-4

C C Q4) 00 C a H

>4H-H

C)Hm4uH,Pa

>4

H

. Z0Hz-II

z

o w

P CQ1% >O~ cn

C,

0 :0

mr04

>4z

0C

-

---

--

Page 96: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

)a Of oo o o

o o 0oD. CD c--*con - 0 t rt r

t·Q Op 01'

totCO s0CD =¢ L0ct OtJt*-

C 0Di o X 3- ( : - 0

> 31 ct D O M CD 3I r > : x rt H. (D :3 e oDX o ' Q co £ 0 W m W n s X

.... o Ct * O 0 .... .... 0 P r"

n O Oct~ toD Oa : pi o : CD c 3

CX CD C 0 0 0 ;3 OD .C. W 0 H. Co o 0 t tr0 ~ Ii o, "· -- W W - 0 0,

_ _ O 3 0 o C r _ _ .t -- 5 Ctt hi Q H:D 3 _ U0 1 - 3 JJ FC

U VI * C co 0 C -. DJ (1 D ~ : ,.,

,, o C p i -i ~D A h t ot, 0, O

CO I Ut o w w Ln O ( D

CD" D ,o C0£ P o H- oO o \ ~ or a 4 O O cDt

w > -n, h.W " CD ,

C) VI · r vr o CIo IQ Z0 e 0

D i i , ,t ,

0 C O /· C D 3 . ,t. CO CC OX CI

rt 0

I m OQ : O bb. .(D.OrtOZ C !COO 0)ZO0 CO CD0) g 3 ' B 1 i Dc- *t t X e . 3 I1 0 OO CO rH< qC r0 .

ct > O t 0 0 .

0 tD r 1Q

0 cD c C C 1 2

r .' 0)C)

y

0

'0

:

10z1.4

'0

02trJC)

0RaCo o

LO 0a c

30Ea o-a 3q a3 Ctiz a

r)0O

zD32OH

MW

'0z l)03 :3'0a CXXm a

HvC)

-i

HL'1aH'-3c4

tlf'3'0

H

H

H'-3F<

Page 97: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

En 0 M . > 0

() ri' o 0 -- n3 a 0 0 n

0~~~~~~P H-. 0 co0 Po

00 O 0- 0 0 , 0

.... - 3 ~ ~' -. ~ 0, .rn Q- ~ . , O- O- ~ ~t pi P o En CM C) X Po n1 3 ~3z

5-u CD W <> :y 5.t. h, .- 0 ~., := O 5- 2 u 0m O O "t a ' X < I' 0 m LI D " 0 0 ) C oD 3 I: I CD H. I-- D ;3 3 ;:I 3 t ,~ >CDX u1 tD O tTJ " ~ 0 CD tn 0

3 tE ;tD3 tQ (D PI e~

_ _ tri o-t (D 3L o a r t (D H. * H.

ooo."0 o m :3 , L.) r_-- __S. -O ~ O I t En in ) co tD w =D I D CD t- - Q CD 3F

.1 0 o: ct 3 CD > o C no u (D tD <

w~~ ~ ~o ·en C o o : *.

-o - - (1) H (1 0 zu 0 0 ~ W D

~~~~~~~~~~~~0 m.% rt 0 > Lo m- 0rS~~~~~~~~~~~2 ft %. >o < co co :r

. D , , , D.~ i ~ - oa , D ( tI ~,-, 0o 0%p n(Deb (

V~~~~~~~~~~~~~~I VI !. CDi

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~t,,CD

M-QO Cob

tu1 m 00 a r

.a .t

D0t 0 c t0

0) (D I (- nDX

E03 1-n 1- 3j -:X<

0D Q ~: zn (D D 0 m 0

H- U jZ H < < f eH £r 0 D w. m 0 z " m

CD c 0 0 rt, t t Pi -I - W H Pt .) 0 . = X

co zh 0 Cr rtQ PO ft X ,En NfDt; ) m " H.

ro Pa 0; D -M )- tm ct

3 nw t: rt ct cl t co ;u H. 0OD H- c: co H- W . g:

0 ia z 0 ct LaQc. ) -- -, H.t.Z

III

0

f3PO:E

Po

z

za zq

ae 0K :q

~-

0c~0:_Pa>1

Hmcn

g02 >00-3 Po:> to0 M=13 m a H

x

>0

H

0XH

HF3a

10100H0

tz1-4t"H�-a1.4

Page 98: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

W, :C :: 3 r:: = a)< X o 0)( mz wm >Xo W 3 Q; £ Q Po t W ct zmt /D 0 rt 0" rm 0 Elw('1 ("! t -'"'rt 0 /1) 013 rD £ ' t 0 m 0 W mGO ~..., . o x £0 o S3.

Q ( ("2

: 0 c::~~ 3~o : 31U:2 3Po 3 3Pom > C 4 L4 X C) t:= C~ ~'IO 0) k--: D PtJ 0 (D !t:F 0 0) I~'

0 W C2 : X 0 o t ~ I0 I>... =: m> -- £ .... C :3 -=0 m ~,,~ ~ ~ ~ ~ ~~~~~~~~~~~( re(l C) rtm~-, ., rd" II.

_ _ < r y ~~~. < :XUt X ^3> W XrtJ fD 3t H. En El O s1- e t -0 m 0 -0 - z '--

O,4. , , 0 W CD ~ W M O-' 5-- -- >o oJ ,l 0 - J -- J 1 " r

co0 o0 hJ Po ho boo CD cW W I O QJ I _ W W W _ I__*s~~e ~ > a __t>;Xa> bi 3g,,,) bJ ,,q o"~ tD o to rl c r't'

oco c co I N) oo D co O-s < e ~~~~~~~I K) * ~ WOP C) 4 I~~~~~~~~~N ICDS I I 0 -

to -J U O 0 at%D O Co NJ V -

(1)

,3 , W0,--] " 0 m r

0

m z rn zY 0- ft oq W'lHP. ( :r 0 D ft Piy (1) H-

).. M :Q'th r ft . M L - L 3 ~:~ :3: ,q (-t-~ ~-Oct '- S 3. C < (D - 0 0

W. F- 0 OD 0 0 DJ . Q"t

O re 3 ~ r- O re : · O'J· * ~l~-J't'l~ - · Pi C W Q

0~~ ~~ 0 0 ) 0

C: ~ ~0 cD 3 Q 33 0a ~- r 3-

co0 0 m ·t rt c::: i s cn rt < < 5 I3 z Q' t H (5 H. , .S ...

3 D -J tD £ W m 3 wco H- Qt W IJ rt £ W Qt

0- lao m t La 0 0 m C

m i Q D tD Wt m O Xe 0 WJ rPt t W

3 ctX 3D W Q ~

l~~~~~~~~a .~~~~LC o E

3

z

Po

zm 0

z z

0 0

txnMn 3

>t

H

trEn0a

toi

C7Hx3

he Po14 :0

0-3 c

3x 003

10PoLAH0

WHvH1-3

Page 99: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

:~~~~~~ ~H ) co 0H t H Z 0 0

0 £) ou~ 0t 00 . 0 0

l O ~ ~X O z " ",

I t? go O

~~~~~~En0 0 3 . _

I o M 3-9- P t Ln - 3 I z = co H1-D --- OD 0 ~ o1 5" D 0 . 50 (1)I -Ct 0 3 3n Q O Q 3 -

I S- ~ H. M- 3 t t En > 3; C Z =5 or OI 0-,-' o I D 0 X x F ~ > X I" 0 " b1)

co e1 .. S- 3 b l 1 ° *-3 tD l3,

~~~~~)tz co ( D Ln :< -__^>_;

ri' el' ~ -'((1 .-, ( 0 l 0 r't

(1 ) rt~ ,q 0 (1) H. .q ( ·i (MI- " = En cx £ c t N) Ooa, 0r 0 CN ....I n :" 0 ) (D <i aN t- z :c MC£*-*

I 0-' O r' - to ) 0 rt ' z 0 : ~ O A

1 c - _ _ 0 _- - Q rW Q W v. tt r Iot 'j , ' t

I ) t VI CD tN a% * iW 1oooc

t. 0 Mt 0j 0 0wI 0 e x st O /so 3. I I | - t i oI

I N N. on &. -s -n i

I VI co VI CD O CD W

I~~~~~~~~~~~~~~~L 0

I ~ ~~~~~~~~~~~~~~~~~~~~ >

O z

En u?0

H s- Z o 3u £D M 0 pi m X

4 to er r < Q: PO 0 C; (D 5 H. ws :3 a Oy r

ct c < 3 el 0 . ( Q* .3tr tD Oa P. :3 w tD :$)=

M O ct 0 t O M O (D 0

H Jr~3 c . p.0 M

o - 0 wa to 0* fD ol 3r * m

= X X r~~~ ~D I

IIIl

3

z

Po

m OQ:0

Uz

0 :U

En

t C0

3

03

0

0

:1-tozHm

z >

>

z z

D H

- x

z

t

VnH

tz

Page 100: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

:D|rt- m 30 3 O 0 30 W W " rt ~~~~~~" re" c

I . P1 -O . t O -t / I

:IX 0' 0 X I

~ UZ D x t n P tl UZ 0 ID 0

OO > P O :X En ,n r O .PO P O co En I.· w - Lo-0 3 J 'IO I~ H. o 3o~::~ 3~/ 10 I~ t~n 3 " C Po . 0 W O-

~~~~~.. s. * - 0- O -W )Q>*

-0 ~:) - to Ol., O (D 0: -, EnD/1 :O & X Di PO W " n ~ IO0 o ,. - -~nD W - ' 3 :. I ~90 MX 3 (D O tx ct w>L M _ t C - n : _ _ _ t .I

O P Pt X X O N , . o /O

· ".- ( t-o-I rt ~ 0 I.~r -i O " H. =- L W CD 0

L" n D to <r t V. Di (DW W - tj (D 0 Ln - XI t >W

I M O W 0 D o C : ·CD I 0 -- , W 0 -· I

mb b o X ow m

/1 0 3 ~ 1& . W 00Lnt-E11 M C) -I I i r I

....... D 3 i iZ w H D t- La Ln taj z :P F~r

(D rt t ct wQO 2r

rt MD~ tF D 3 s~r

3 H ~~~~ ~3 33

3>D z s ft ft m

w m m W ct ct Q o (:O

CL ~ t 0 3 O W< H. la,

3 3 = $- (D: Q 3

i = W O WI z (a ct cu.

l H. 0 0 n 0.~~ ~ == .A _ H, :t ._

z

:z

-3

O O

0t~cicoEn o

En to

13 13

n03

zH

z >1-3 00> e0

0-3 SZ

D H

>1

HO1>aWIHr -1H113k

10POt7lHOWHt-4H1-304

Page 101: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

0 HEnO H 3 C 30 l 0 0 I 0 Fs b t - I t r: r. 3 c

0 ~ ~~~~~~ XI O 1 n £ - Ft X = r~~~~~~~~~~~~c c Z

m~~~~~~ I CD 0. ('i m D1 0

I. . H. 3330 0 0 :> W En w o- O 10 z -jP o I

:F D= Dt,:r1(D I1~ ::: I1 C' ~'1 u) :::t I~ o1 , 0 0 ct H. 0 II 0 I1) ct " 0 51 L.), , TJ0 0 H- ~.--,Z:1rr t'~ :H- :3 m n· :$ tr ("o - : > Z) ~-...to

CO O' . t -. m .. 'X .... n O

'~ co :w '" u 0 I1) co o~, oft ·" uC-

I) CD- 0 tri~t I ~ II, I' Al(c1 " CD I 0 I-'3 > -i > 1 i- c -

I~~~~( 5D rA 0 5D III z III cQ 0K O ct 0 :c Q ~ tQ O O co OD 0 O 3j tj C HD

1 ) 3 5 z 0P 3~ ~ 0W -3 ON -(3 ~ to-i tD 0 = II %O % D 3 X co CD cL X - :0ttD*- DQ ct ctI - I t I. N c I I. ... H-3*-O O

tj &. ov :; o bi -A¢ to Uj Ujt o co co >o tt1 m 3 mbDQ oo<

eD Q ul ~ ~o LA OD 3> co N :ct O W3U co n < P. tD

0 3I q O ~ ~~z 0 £ O

ct

0|~~ ~ ~~~ ~ ~~~~~~ w En 1- ct 13 1h H

~.~~~~~~~~~~~= m. 0 {1 z I~ ~.DI~ 0' = O H Oct

I · , , M .·I ~ ~~~ ~ col ~ ~~~~~ -- n Se

I-d -. o -.O~~~~~~~~~~~~~. H. O .2

~., . * . = 0 m) .

0 o 0 co · ct QI b > H-0 X > bI I I O y 3 b ct Ot

1 3 t) 3 tD v X £ ut (D 3 (D~~~~~~~D 0

III

0"0

Po0

tz0~0,. -

3

w

En

( ¢3 O

3

3x

H

En

z >H Po> Po

D7 H

t-H

0

Hq110

3010toL-4H0

03Ht,4H0-3

Page 102: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

D" ~~ " ' :uM)-- : :r ~ ' ,^ uW " 03 x H 3 :o En 1"0 r , Z " H 0 O · -o,

... ~ c ... :r .. ~ ~-c x5 -. 0n ct CD v 0 :0 ct s:~ 0 I~~ ~ ~ ~.-,oO n ¢

--. O t t --- re f~ -- '-r - -~' (1)0 ,r 0 3 O CD

c~0,o ~ . to.. D o ~ to ~ >> O 3.. Ia~c c D' 0 . 0 0 n ' U 3 f %D b... > CD (D.. >.. . :3 0) . .c t Ln M . .... * z O co c

O ~ M o G H-0 c 3 (D t 'D_-.,o 1) rv(tQ 0 rvt . _1_

CD ~ CD,. ~0~- --~h ~(cv n'Xo o Oy -J -,- 0) ' a) y X PI Z

O z 0 - > X 0 la,> to 0 · 1 $.O- .

ODLO r D " _-_ 3 ~ f: · t O "O -J tt :5 tU ~' m 0 OD 3 to ONON C0 -kD _J X t· O M 0 O w I I L1 Q I I *- I ts e I ~- 1:l%~mO Z )'0 t0 0 z L X W

s U 1- I-O OD C) zJ X s1 P H.J -1 0CD w o ft J C o

4 I VI: C) c~= ON=c' i no I c' o '.t H z 3 KnH .P m 3

o w 3 0 7 (D " ct > 3X:3 em M O tz w e -

trD 41 0 : w 0 cn P' 0 M o DO (D 0 c - ODeoD

c t rt O y Sr }f*) 3 W 0 :; 0 a )

n O 3t 0 H- < (' O .

W2 o tt 0t c t 3 tn 1-3 rs LO (D3 a, b z

fD ZYD > ) fD :L V mD z O .J= D M= m XCQa2GDct fD CD t I. coU

< m 0 Q co H. 0 (D 0 0I~~ ~ ~ ~ ~~~~~~~~~~~~ ct m

- . ¢ l ~~~ ~ ~ ~~~~~~~~I t rt

1I

3

z

It

"0O

21 zZToO13

n0En toco 0

n X

,3

OX:x,'o

Z

i-H

10>z D-X-4

> ,rit~iOC)Hx

HO3tz

H

Hi-i-i,-]

Page 103: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

S > n ° £ rL 3~~~~~> E Ol< 2;th0 5 0

* 0tO ~~~:t O tO = O~~~ D " WD PCX> =>=

2 OtQ- rtc 3 < toz X D tof O ¢~~~: D =>O

0 ~ ~ ~ 0 ~ ~-o~ 0 ~ ~.~ 0 0 ~.

Z D · < 0 D Z M r O o D P- =. w X.MC- 0 5 3 3 .: 0 > .. a .. 3 H t.^O 0 a rZ t tC CD M :^ rO U. D 3 rt : tytV

(0D Z - W - C 0 CN O " H - (D 3 O D 0. Po El .o PI 5n pi Xu O D' 3 fO-

0 0 w w w O W 0 - 3 t n 0Q- O t- ( -> _ D C _ Z 5 ( D En NJO~-F4' c3 ;v -IV x > 5M B< 0Q 0

I to ~ W 0 W OD w z t^ 0 WO 0 ) H. 0) 3 O tn Ht 0 DR DF-0 X(D O t O CD MO C>tS m tQ N) Ot tj * ~ 0

|~O O: O~ >1 o

0 0 3 W Ca% >D wO £ >XC _ 0 rt 3 3 (DOD X (DO EnD>X 3 CL ~ ~~~n > > O e O tt O y b X W X >~~~n q

= C = =D 0 tD Po z . . .D< o~~~~~~~~~~~~~~~~~Q to 0 Fh=

Q VS QM ct ; -

_ t rt ct ct nI I I~~~~~~~~~~~~~~~H

,,,~-

IIIl

O

3

Ht

>0-3 00> e0

m D

M XH

xa

H

1-3I

Page 104: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

Appendix B

DETAILED CHARACTERISTICS OF

PRODUCTS AND APPROACHES

Massachusetts Institute of Technology, 1989. All rightsreserved. Reproduction of tables or quotation form themprohibited. All requests for grant of permission should bedirected to the principal investigator, Dr. Amar Gupta.Telephone (617) 253-8906.

©

Page 105: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

- |

l-

0O4JvlCcnC 0 Oz

a

1

4 r 1 n I

'w 3. N Ln Ln

I u Ln

a C to o

S4 04> 'tT0) 0 u) ) - -~ e Ox c

> -a 0U,C)

m 0 ...

m EI Ca ec vI E oEH V .- O

4O O

0 m

coo'

aNU_

oto

.,-. oC ak : ~ n· r, O~ OEE cncor-... u 0

t( 404 co O C oC

~0

.H

l:: 0

-H O 4C 0 r= Qu E

< b *H U O

Et m cn u:3 CL

O

,4ak

C OVU

a -.,-I

E

Q) £ _rca V VI

0 N cosq c

Q)0 Uro o4 O .o· , or - 0o

O I > E ,"~¢ O N M COCW Cr Q\Q

..~ ,-. ' ,..co 0 Ss O U U:rr * rl O OE k C X1 C 4OO Q W L_3: C H _

; : 3 mr e C*,¢ Q.;C a Sn E O

< Q Z 4 n U 04

cn>I

.-CP

0Q

ml.P

,-

4 en

a V--I CcJ00=

C co

4.) r 4 >1 .-L4 C

.. l 4 0 Oa o : a

0.,ll a.4l)

0 0

0 04

4) 0~l iII c ea O Po .Hi b

CYW eO ;

C Hor a S *

a * o 1 X~ a

O

,i4

xO fO sH

Q oC) ,,

l-

111

Il

0zS

c c.l,

C

C0.1-

30

a *.-a; C

Ira

0c-Eu

· 4

a 0rq

t't.J) 4Q)C

o:; 4C-QZ

. )

4i0r Flo> C

X (V

4JV I

Cro

0

.3a £r4J a

o c,Z *H> to4 coW a$4

(II4J0 m]

Z .,q

ICH

w

> i

£I '

Em

crO 0 0

.

a _

$4 C0 4)*[email protected] l'~ -J

II mC n ·l m l

0~

O]4J 0 13~. l CC C: 4w CPQ)

4) a) 4oJ e

z s

ma w

4 m w

Mtr 4- z0 w m 01 J PC 4

£~a

o

0*) 4

4 c

C (E 0.4

0 .

O H

O0 5H

-r4 $

H

4H

UHc7

Pi

:

>4

CcI-4H

Z

0$4

C.)wInd

44

La

z -

u

P X: wa E-

a U3

t0C/)W

U au a

:E4

Z 0 z

0

4

>4Z

0

Page 106: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

n.~~~~~~~~~V

-~~~~~~~~~~~~~C ( C-(' c n C: C* ~l 2

Ch.~~~~~ ni C'>

-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~C -eV cor pt P, c-r 1 I a a I a, So CA a C. -

::~ cv C " -l w ' f 00 " .. D ~-- w ~' CD (:I a P, 1 3 P I amr· I r~~~~~~~~~~~~~IV a o V Ca or I a C I ct~, C- cr -c1CDC. V> -c r

f ~ ~~ 00" rIfI"' ~" z

~~"

C P t2, C-D' (' ' - ,- 2 CA l :r· cc I a I a p, 1 5 a I c~~~~~~~~~~~~~~~~~~~~~a tllala pr 1 I a I P, cc I aC, M. > 1

II~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~I:

oy co0w ~~a

C" w I3I a 5 3 ( eV In (V M M (ii

~~~~~~~~-

a, co c eC, w n = a o C- II =Al =. - - CD~ III

a re L~ I 1 L9~ Pr 11 nCD c~~~ c-( I a I p, a a ~~~~~~ I II Se) a~C

r- mC3 aI ro t a acnw I ii to n

a~~~~~~~~~~~~~~~~~ w toIaI - II C, Pa I 3 oC

TO a - =r-

a c cc a a I r· I a I W a~~co c

co - MC. a, C InI c

M D n -0 I ( n -0 1 1 t Ca P, cc a I r 3 3 I a a a 0 eD

ru a n, r ru a cv 1 a, ~co 17

m P I~~~~~~~~~~M C CD M t- I I a Ca C> w ?C C> C >

a a a CO ( I a a cc nr ( t p, C CD C:, szc* a r~~~~ar I I Cp ~~~~D > w S I C = I P

cc cc r- a a I I a a " =C. C S= Co I O Ca a cc CTCCI f L rS cl ) o t I= C

cO r~~~~p, a I I L cI-, c a C "11 D

I I CDE: CI~~~~~~~~~~~~~~~~~~~~~~~~nCP I CL3~~~~~~~~~~~~~~~~~~~C L

I~~ ~ ~~~~~~~~~~~~~~~~~~~ I Mrrr ~~ I V CT a c- ( or·C o 3 I , eC~cr ·- j ~ CD

cc a a ( a P, 3 CI a ) I a cc c--- 1 a r· a ·-- a a ct I a~~~~~~~~~~a r·~I cca Caa a ta cc~ ~- 1 acl a , cca 3 3 2

a I I Pc a - I r' - cc cc 3 ) a II~~~~~~~~~~~~~~~~~~~~~~C Cro c I 0 a cc I a - r· I P, cc c~~~~~~~~~~~r a a ~~~ ( II 1Ia I a Pa 00 C ~ a C1 r ) a Q) ~ a a a , I IC

3 I a 1 a ~~~~ D, a I a c-· a ~~~~ cc c-· I II a~~CSo cc ) aa a a I ~I r· c--· a I n, I IV

1,5"

Page 107: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

330 1 a o I ao I I cc c 9c " M

~7 > 3 C'2

I > 1 I n

t7 C

IC

-M -0 C 0 -M M C 0 O1= CICD

C

C" ::

I :

cc p, I a 3 I E; ct I LT)~~~~~~~~~~~~~~~~~~~a tco I K D- C w CD IV C

I I .o

C

cc I~~~~~~~~~~~~~~~~~~~~~~

=PI C MC -C c a 3 c- c C o c a a:

C)Pr a a P, f I,,,

me ,.el

- O C> C CD " ,0

=

C/

C,

I ,

wt P C, a 1 a 3 w d C6 t a a P

MC = = .. ~ - , ~ c-- , =, . - =

cn- t -a ~ ~ M.- M C o M CD ~ ~ ?.,t2

I~~~~~~~~~M

w,

~:"

sm C

It~ ~ ~ ~~~~~~~~~~~~~~~~~~~~¢~ g~Irpr,

I

~

3 C

,.,, ¢,-- g. c ,- '-~ P.

r! mtr,- 3 ( Ct' Wj I ru o~~~~t~ C> o I= - - =.I r r : a " 'I

e I v MC>c~ s C. C

-r C> ( C I am O c V C.G

e C2 = 0 Ci, M .o a 1 0 , CDc

CZ Pc cc 3 I rP vJ ~ a 0 .co3 cc cO cc a ) C~~~~~~~~l 10 I 3:W I s f~~~I, c t c Ic- .a

~":J i

0 ,-I

C3 p CD co

t v t e I.C ~ 310 Cc CD c· 3 I I Pc a a rD I ~ ~~ ~ ~ ~ ~ ~~~~~~~- c

a r· 0 a~~e C I ad a1 P M

I P, cc C5 3 a ( H a a i a : ¢"*'

W

a 0 a rC ( a a v C DI 4 %=( -C

3 a ZS~~~~~~~~C ~ = )C c r a, Cn,

Io-·CI c

M C 3 ~~~~ 4 C O c o t o t= ~ ~ ~ ~ O r a m rr, a t a ' 3 cc t P, ( I N Q)~~~~~~~ l~ ~"- '

SW co nI m

C "- EN Mc-c1CD cr c a ( cn I

a t

co C, C. c I,eD~~~~~~~~~~~~~~~~~~

j

CDI-.* m w C.

a~~~~~~~~~~ ~~~~~~~ .·C F- ~I . :B f c m r ,'~ 1, .. .. . I I ff

a, CDC3 a I a I cc C rD I CC m , Da P, CD I o I e o 3 1 !a n CD S

1 0

Z .

--C' 0

t~r I "= ic:0 C CO f

to -CD I f

6 . C- a .-

C> M M 0

c I

,

CD Zr, -.4

P,~~ ~ ~ - V- ° i - a a '=cc C1~C f I

3 ~ a I Oto Co r

En C o

I - 0 On "', l

,I -

to M- > C D

--,

m"

3 I I

P, I a - cc

C C> C C-

p a cr~o a ) a a w a I r a ( C- CT Pco

C-

(

o ~ I a I 3L~'" I a I M 00 coCu I sP r · O C CD I d ~ ~ ~ ~ ~ ~ ~ n a m c C C . CCD o LI C ( 3 a 3 ~ ~ ~ ~ - I C~ a I OD C a , - , C

C,, I cc 3 1 Pc m serm r I ta c t 1 3 a I m C w I P P~~a-CD , C D , r

Isrp I

I '-",~ ° "M ""' CD "-

C. C.1~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~r · cu ~ ~ I n , c ~ cu a 1 r~ la

C " Z I D I V

3 1 co~~~~~~~~~~~' 12 al,

r"~ --· , o ~'2 OI - -a~~~~~~cr ~ ~ ~ ~ ~ I=~ ('

a a w i ct I ?D CD c'

- erP, I sPI13- o o o a o c - i 0 1 ~ C i

III

0

C12

C0 t0

o

0,-II31

C=

~mOaT

=C~00 C0

ee0-C> so

o

Page 108: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

tl 0 " oIac |~ _. . O o o o

~, | n

41 n I t= CZ, ~ ~ ~ ~ ~ ~ ~ ~ ~ ,

r 0 W o · I I

CD c+ I o? I f Ile I Io a I I , C .t e c _ ,' .

I DJ O I~ 10 f f

Ol C ~-." O C :!: O=. k.,.cnz~,.~0 , ' t l ( -"C _ " _ ..C0 al M

0. C-... P - C: >" :10~~~~~~~~~~~~~~~ 1 gb

| t cr _ w O i-..4 tO o ~:_coW b o .e ob

I . a, L to I C.D C> C O n " "O =r M C, CP

I mL- C D IIt~rI M I I rII . - . , =,C.I

s ~t c t:DC

§ _P0 n ~~~~~~~~~C_- C- c et, to:MO l M . C

C- > o_. co s o C2 1

r~~~~~~~ U n C-, C._t~~~~~~~~~ t o , -, o ee ce =X zt02I U a~=lI 3

.zI -

ol X _* eD o I C.: U OrO VZ CI 3 D r-r f o CE oIa Cc r _. N =s C _- 1 - j _ fb f I

! ~~~~~~~~~~~~~~~~~~~~~o CX r IrD 0 X I m n> ,1a ( cr n, a C>,, cr f a o o l~~ Pa a ( (D C3 Pt I 3~~Mc4 M e~ c C

,, .CD CD I (O h) O I % C> I C

t o | C I X S G W

co t c D

C - C t 1 _ j o3(

I P al~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~I _C cm @O a Cc t a O a a 3 I cD ~t C=~ LO Cp, ( o, m caI 3 I~~~~~~~~~~~~~~~~~~~~~~C > lcn a

cnI 3 W I u~~~~~~~~l c~~~~roclraaI er·~~~~C r

'IcolI 1=- :

kl

I ccDu Cjro

W L

O

o CO

cZl,- IW11

l ClO

CocOC v1.=a mCI e

Page 109: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

III

I ~ ~'~'- ~o I='-1 1cD CI 3> 1 cc CCI2a o I0303- 0 00 'On 0 II 0 0 0 I ( 0 0 43 C>c

ebi

I _Iit ~ , .0 - 'V0 (00 0- 1-.3 ( O ( 3 C 3tn M 0 < ~0 (

C-A :3 c · o w t= CD Ca l 03 - '. co CDU2 C I Ia co a rV03 W4 I ii I~~~~~~~~~~~~~ii~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~I ~I C~ L' C- a~ , -4' C , C) m., 0 A C> C 4 ¢'II cr ~ I s 0r 'V N C) (0 b-I 'V 03 'V a O -3 - - '1C D1 - M oil if

II P C~~ It c r C> a CD 1 CD(D 1 c)1-C C 0 I0 C> -m C. I* (0- * -.. 0 1: ) -II - I 0, v. C 0 C , - - 'V U P a a 1 II=> C. CI = = = C . r~ C>o~~~~~ - Co '1 C V , C~..'- - Ci i c ( CD -03 0( ro ivr Cjr -5 - Co 03 c 1 -0 D K C -- c 0 03 "I c. C . C° ' . 0 ! -( (0 C) (0 3 ) C) 03 Co a 04 E l

II 1a N31 m o - a- -- a II

II , -- c o o ~ 3 I c a. . -,~ ~ . C,

" ~ ~ ~~~~ " C> "'I W C- M~ cn I C'> co!~C3 .0-I...)C-.4-,C) 0 'V 03 0-C) Cf

M CY C~ I a,11 IMS I'm C> C> 0 c0e - 1 : u ~ 0 ( - >,, CD ~I 11o c/r( al c-c~~~C 0-- - .0I II ct CD C> ar~ce a 01c , t, 4 CD 3 a 0 1

'-I .- ) ...* .-. -- o ' - ~r c 7 g 0 ~ C l::::s ~- . c--

ll' a ID P, C° ·r P . . .-~"~ =- "° or cC 1 P -t . .. , , = I-II o * 3Co/ u C . ' 9 -b CD C. II

I C C" 0 "= 0" -mt 4 .- (11.0 0 03 0 i iiIt 03(01 -( '1 0. I r· Co1 o1 a c CD r· n, I 1 II a I P, CD

9. ,- - M ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~!. C>~~~~~~~~~~~~~~ M 'CD m- 1 . CD ( n V- CD I- MIC /a E: a P, ~ ~ p·C, IC " w f- C- C

-- a Z -.4 D W 04 M C> If(0 0 0. C) (0 (, s= a 03(0C1 P ( . C " - cI ' 03 0 CD X > r- c' M'a.(0 (0 1--U 0 1- 03 Iq I ,a c- 1I C- 0'0 0 0 1 c

03 co C>- to - 0.>0 0. M0m C)c I a, M-0 C- (0 1-4 '-4 0 IC) C, 0. Co C) C, W (, c '- 0 cc ( 3 (0- CVw ' -C - (I -, cn I I (Z 0 C0- I-S 'C 0M 20 l5a a C-3I- 0t -' 0.- " c 3 ( (0 03 I- C- I- 0 IIt a ro t p, o tw 4 31(0 C M C> 0 30co (003

I= " !t. . M C, C- c a I c t C C a I= C, w a c c - , CD P a *a1 1-1..)*C) S r40 0 03 C

p, 1 Cu 1 a- = c ct C> co C C> = Ct- C ', M .0- - , . " 4 M = (2 0 (Z.. - 0I 0c003 03 ( b- C (0( '-03 3 0 -ii 3 C9 0 bC)0 - 0 C W CZ Co - (0 Co y-'CA n. b 0 - 1C" '" o c- 2 '1 1-4II~m (00 I S Co 0 000.03 (0SC o0-( I-'j4 (2 'S 1-I m. 0j (0 'V~ ,a a ,

M C Co 0C3 Co- I co C. 0300C r'oa ,, - I c- 0 0. - __CD C>

CD ct Pracc(~~~~~~~~~~~~~~~~~~~~~~~~~~~C

CD t=1 IwfPI LI D P , IICD .a n w Vo M C> wii ~r 'I ___t c p r a Ic (0c 0.' - . -03 0.I- C~ r* 03uP, IrCD C- co Co ( 03 cn C-

I ______~9:. M wT =I t= 0Co0 03 C. 0 ' 0 (0 Co m >cC C 3- I- c 0 Co b- 0n '- 040-C. C. 3 *a (0( 0 l C-I-C>CtI I · a c - I ;~a p a c )C C · acM 1= C.ID- 0 0M

MIc c P ( - cn cn aa 1 -·C, Ir P r 1qc· :n =c - XD (D a I~ > ~ ~ r Ic

n, ro CD ! 3 r--I tP O cq c ,I ~ c -IIa-' 0, I I-' 00 0 C 0.. I ' g 0 'V 0C II H - 0 o s (00 IC I rCo~ 1- l C) 'V - 0 a1 (2 C I I 0 M (2 -5 C2CD ' C (0 0 3 I (20 l- S IC 3 C) C o 0 (003 0I l-. 1-- l- C o I a 0 03 0D '1 0 C-I C1 'V 1-- C) I- Is~03 rIrI 0 -c a a I, 03 . 0.DC 0, a ..ci2 03 .-- , C-'~ -. (0 1- C~s - I (2 '1 03. 0 W-.--. .. . 1 -41 1I t (0 (0 (2 0. 0 a 0 03 c-, Cc C o '- P- , 0) - 0 -.Ila ID 03 er c-- CO 03 .- l- -51-cta - - ) 1-* 0 0003 5(0 r C (2 -0 0 . 0t C-I 04 8-- -J m o I-Iv , r a - 00a II O Z- *1 naa D~r r c I a3o~I(0 1D O tC~ J I 1- Co 0- 'c -' . 003 , C) 0 '.0JI, '1 1.8f0 1- 03 IC Co .-0 03 C -I- 0P I 0 - 0 Co LII0CO 'V C-I (2 0 r a c rr c- ·(0 0) 0. 03 0. 1. 5t C)II '1 '1 1- (0 c Sc 0r C-I rI- - (2 a n 0 -.h C- ct ,1 Co C12 'V rI-0 D 3 (2 1- ~4 l- 0D C)-J 0 - 0. Co 02 (00 i 0 * 5 ICo( 5( .0 00 -0 CI 0 )-

-4

M

C)O

toCCC)

C) co

,-41

O -5C) :(12

C)

(~C)

).

i-

-41Cd) C=C=

M-5

Cd)02C)

O, C

Page 110: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

M.I.T. Sloan School of ManagementWorking Paper Series

Papers by Stuart Madnick

Paper # Date Title/Author(s)

1839 11/86 "A Microcomputer-Based Image Database Management System,"Prasad, B., Gupta, A., Toong, H., and Madnick, S.,

1776 4/86 "INFOPLEX: Research in a High-Performance Database Computer,"Madnick, S., and Hsu, M.

1685 6/85 "Impact of Schedule Estimation on Software Project Behavior," Abdel-Hamid, T., and Madnick, S.

1366 10/82 "An Integrative Approach to Modeling the Software ManagementProcess: A Basis for Identifying Problems and Evaluating Tools andTechniques," Adbel-Hamid, T.K., and Madnick, S.

1334 5/82 "The Dynamics of Software Project Scheduling: A System SynamicsPerspective," Adbel-Hamid, T.K., and Madnick, S.

1333 5/82 "A Model of Software Project Management Dynamics," Adbel-Hamid,T.K., and Madnick, S.

1047 4/79 "Properties or Storage Hierarchy Systems with Multiple Page Sizesand Redundant Data," Lan, C.Y., and Madnick, S.

1046 8/78 "Inforlex Data Base Computer Architecture - Concept and Directions,"Lam, C.Y., and Madnick, S.

V 993 5/78 "Compostie Information Systems - A New Concept in InformationSystems," Lam, C.Y., and Madnick, S.

920 3/77 "A Systematic Approach to the Design of Complex Systems:Application to DBMS Design and Evaluation," Andreu, R.C., andMadnick, S.

910 3/77 "Trends in Computers amd Computing: The Information Utility,"Madnick, S.

883 1/77 "Institutional and AD-HOC Decision Support Systems and TheirEffective Use," Donovan, J.J., and Madnick, S.

846 4/76 "Virtual Machines A Concept That has Comparative Advantages inSecurity, Integrity, and in Decision Support Systems," Donovan, J.J., andMadnick, S.D.

Please Use Attached Form When Ordering Papers

Page 111: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

M.LT. Sloan School of ManagementWorking Paper Series

Papers by Stuart Madnick

Title/Author(s)

"An Application of a Generalized Management Information System toEnergy Policy and Decision Making: The User's View," Donovan, J.J.,Gutentag, L.M., Madnick, S., and Smith, G. N.

771 3/75

770 3/75

769 3/75

723 7/74

722

648

645

5/74

3/73

2/73

"Data Security and Data Processing," Madnick, S.

"Infoplex - Hierarchical Decomposition of a Large InformationManagement System Using a Microprocessor Complex," Madnick, S.

"Design of a General Hierarchical Storage System," Madnick, S.

"Virtual Information in Data-Base Systems," Folinus, J.J., Madnick, S.,and Schutzman, H.B.

"Application and Analysis of the Virtual Machine Approch toInformation System Security and Reliability," Madnick, S., andDonovan, J.J.

"An Approach to Information System Isolation and Security in aShared Facility," Madnick, S., and Donovan, J.

"Recent Technical Advances in the Computer Industry and Their FutureImpact," Madnick, S.

Please Use Attached Form When Ordering Papers

Paver # Date

776 3/75

-- ---- -

Page 112: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

M.LT. Sloan School of ManagementWorking Paper Series

Papers by Stuart MadnickJohn Norris McGuire Professor of Information Technology

Paper # Date Title/Author(s)

3157 6/90 "The Composite Information Systems Laboratory (CSL) Project at MIT,"Madnick, S., Siegel, M., and Wang, R.

3180 5/90 "The Elusive Silver Lining: How We Fail to Learn from Failure inSoftware Development," Adbel-Hamid, T.K., and Madnick, S.

3119 2/90 "A Polygen Model for Heterogeneous Database Systems: The SourceTagging Perspective," Wang, Y.R., and Madnick, S.

3120 12/89 "Motivating Strategic Alliance for Composite Information Systems:The Case of a Major Regional Hospital," Osborn, C.S., Madnick, S., andWang, Y.R.

3100 12/89 "A Polygem Data Model for Data Source Tagging in CompositeInformation Systems," Wang, R., and Madnick, S.

3084 11/89 "An Architectural Comparison of Contemporary Approaches andProducts for Integrating Heterogeneous Information Systems," Gupta,A., and Madnick, S.

3092 10/89 "Schema Integration Using Metadata," Siegel, M., and Madnick, S.

3091 10/89 "CISL: Composing Answers from Disparate Information Systems,"Madnick, S., Wang, Y. et al.

3061 8/89 "Information Technology Platform for the 1990's," Madnick, S.

3015 5/89 "A Global Query Processing in Composite Information Systems," Wang,Y.R., and Madnick, S.

2075 9/88 "Software Productivity: Potential, Actual, and Perceived," Abdel-Hamid, T., and Madnick, S.

2061 8/88 "Logical Connectivity: Applications, Requirments, and anArchitecture," Madnick, S., and Wang, Y.R.

2031 6/88 "Motivating Strategic Alliances for Composite Information Systems:The Case of a Major Regional Hospital," Osborn, C., Madnick, S., andWang, Y.R.

2029 6/88 "Inter-Database Instance Identification in Composite InformationSystems," Wang, Y.R., Madnick, S., and Horton, D.

2025 6/88 "Connectivity Among Information Systems," Wang, Y.R., and Madnick,S.

Page 113: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

M.I.T. Sloan School of ManagementWorking Paper Series

Papers by Stuart Madnick

Paper # Date Title/Author(s)

2022 5/88 "Implementing Message Oriented Transaction Processing for DistributedDatabase Management Systems," Bhalla, S., and Madnick, S.

1862 5/88 "Evolution Towards Strategic Applications of Databases ThroughComposite Information Systems," Madnick, S., and Wang, Y.R.

1981 2/88 "A Framework and Comparative Study of Distributed HeterogeneousDatabase Management Systems," Bhalla, S., Prasad, B., Gupta, A., andMadnick, S.

1980 2/88 "Modeling the Dynamics of Software Project Management," Abdel-Hamid, T., and Madnick, S.

1977 1/88 "On the Portability of Quantitative Software Estimation Models,"Abdel-Hamid, T., and Madnick, S.

1975 1/88 "Facilitating Connectivity in Composite Information Systems," Wang,Y.R., and Madnick, S.

1866 12/87 "A Conceptual Model for Integrated Autonomous Processing: AnInternational Bank's Experience with Large Datebases," Frank, W.,Madnick, S., and Wang, Y.R.

1876 4/87 "Hierarchical Timestamping Algorithm," Hsu, M., and Madnick, S.

1875 4/87 "Time Slice Manipulation in Information Heirarchy," Hsu, M., andMadnick, S.

1871 3/87 "An Integrative System Dynamics Perspective of Software ProjectManagement: Arguments for an Alternative Research Pardigm,"Abdel-Hamid, T., and Madnick, S.

1868 3/87 "Perspectives on the Effective Use, Planning, and Impact of InformationTechnology," Madnick, T.

1867 2/87 "The Economics of Software Quality Assurance: A System DynamicsBased Simulation Approach," Abdel-Hamid, T., and Madnick, S.

1864 2/87 "A Temporal and Spatial Locality Theory for Characterizing VeryLarge Data Bases," Madnick, S., and Moulton, A.

1861 2/87 "Key Concerns of Executives Making I/S Decisions," Madnick, S., andWang, Y.R.

Please Use Attached Form When Ordering Papers

Page 114: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

i~~~~e coc>~S-c-~b:M M

~~~~~~a-c -vo Pron~~~~~~~~~~~~~~~~~~~~~~~~fea

..o -r~ G CD ,-3fE: - ." -o*3b>c

k,... ~ ~ CD #.*C,* -_- Ca O :r -' mML CZ, _ = " D - toU. _$ a D o O C

O t Q - 0b f C nt n O - tOD t D C a O = a O _ta _ n t

mm O E - - "C Ol~ Cl ;I __________ _____ ____ __ .0 Q '= 1 ". -C Cb -w . a. D M _ 0 -

I:: D - y 0 :qJ- 0 k"C ' 0 r* Di. 0 ~ 0 "*0 : ::r S .. -..-, *" ~mq' C, 0 D

-. '-e0 _ 0- -- a" C -, a, _CD ti 0 C - C

C C- . c _. _ i. Mc b_ _ -_ so C Ca C. <D _ o -

~r' m-~ ~ W ~--' - ~~~~~..d-

0- 0 ~~~~~~~~: ,0 .0., -,- 0 "- 0 C1- 0D , _ - ID I C IIcO 0. _ - 0 11

C c o C 0 . .0 b w

0 a '- D i0 0 .~~~ 0 Ds *-'- I0 o f"~ f~; G~ -.-. D ivid C,: Do, _-. _ ,:_ <~~~~~~~~~~> _v : Ow

C efl~q,* -O 0 = > ;-cv - b rC 0 CO -0, OW 1 -C~ Oi _

(I - E - a.D C D O

:X~~~~C C O IV cO CAl CU a-s a0 Ca C7 D c o _ 0 0 0- D 0 0 0 C O

co 1~Di0 I to r C", r 00 C = . 0 0 - 0 CI- . . -- t , CCa 5-Z 0a U 0 0 Cl ,,. C, 0 M0t- 0 0 0r 0 Ca ta '0 rw

0 r- 0 s0 C- M _ '0 SC D o U C D - C

"~l ~ 0-1 0 C CD -

.Ew a. C A ' - - '-0 I t -'- a C' to C-C a C, -w

l X * C > b Cp *_- >~~~~~~C OQ 9 ........ -X~~~~~~~l S_ w C :D 8 *

rb~~~~~~~~~~ C^D M M D 0 . I

C o 'CS C, CD co , .- = I go t _ _ t Di || D0D 0 o 0 0 - I O _

C - i - 0 C c0DCa 0-'1 > C- t- 0 11

CU C C-, 1,- Di-i i. 0 5 I('Di * -

- t . 0 $ t

C. :s - G0: C -_ tS*_SXI*1 C= -' co 11I C-* tt i tA 0 -C) c o 0 0 - 0- 1

_ r I --C B _ m o 1am *- 0 C 0 I D ° 0 C C .- C e-It~~~~~~~~~- _* o- w I- 0 U- 011 ~~~~~0 5-tC 08 _0 _--r- I e , t

11 _CUZ _^ _l _ I ,. .0.11~~~~~0 '-9 0 - 0 Di _0 0 Di I11 * b fo c X 0 0 0 0- O I CS )

11~~~~~ X 0 Di 0 O-1

li o ~ ~ ~ D | _ saO 0- -_ ||ILM . - - C 11Q0 Di Ca s-i C, 0 0-r-rO~~~C

co-i..

a0 - 0~~~c

ev co~~ 0 s- co I~~~~ r'

0

C)

M o

- Sco co

Cd24 Diag-g-U= MS0

o0

Oa0

Ito 0-coI

I0

_1___ 11___ _X^II( ll . ____11. -- _ _---. -. .

Page 115: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

111

to

ac

_e _- 0v _o _ _ l~~~~~~~~~~~~~~a 0

" " ~~~~~~~~~~~~~~~~~~~~~~c o RC C D

co , . C co - I

X X tD ~~~~~~~~~~~~~~co oc

4 C . C,C~3 W _.X * {

en OQ

(2cofl ~ Ot

en~~~~~~~~~~~~~~~~~:

C~d . toco

0 0 0 B . . n0 I~~~~~* o* '-I o n 1~°~

~ (B (cn (B r O

X co :S S: ~~~~~~~~'1 Z V

I r" -. r t-O .

CD lC> C' _ k '"*:ela =: ." O9

Xo r" o AC M(" B_ * n 0 'C

CD k,_. o' C2 feniI~~~~~~~~~~~~~~~~~~~~~~1 c a W I~ Mg ~~~~0 (I b ~~~~~~~~~~~~~~~~~B Mh o0-|

Is C. crw- 0

Iu ~ t4r t_- I~J S:

C. - . M. M mI n~~~~- 0- -n 0QCX

IfC -. 0

C, co --

v~~~~~~~~

.o M n (aj - . -co - en

I~~~~~~~~ I 0.. _. Bq2

I~ ~ ~ ~ ~~~( ( (B (Bo

(| 0 C 00 en O I

o(B 0 a, 0 0..(B~CW MW0 '0 ~~ ~ MW (B 0CD to

(2 B, (2 (B g~~~~~~~~~~~~~~~~~~~~~~MW a~~~~t, ~i(B en 0 ~~~~~~~~~~~~~~~~~~~~~~~ U~~~C IC~.

g~~ (B g~~~ - t C2 ~~~U ~~~~0 (B '-~~~~~~~~~~F B U

C0 C

0 1>C= cnM .

POOWao.

M

oo- cwDu -

Page 116: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

to to

LX~~~~~~~~~~~~~~~~~~~~~~~~~~3 r' ,-|~~~~~~~S~~~~r~~~~~~~~to ~ ~

. :Ito 0c~~~~ 0~~ICD to .-

to - Cae- C~~~~~~~~~~~~~)d ~ O 0 r : : .- '3

9m~~~~~.

to0D a, 1o O ccc

i,,-.4 cla c~~~~~q- ICla E

0 ~ ~ ~ ,,:,.-=r k,, 0, ~ -' i-d coI O D C

,_~~~~~~~~~. p to,,Ca~~~~~~~~~~~~~~~~D P

"""1~~~ ~~~~ ~ C.P CD' --,CI).-C~ :''t"to CD co 0. -Ca. o"~~~~0

CD00 1

::rI I B c ~ ,ql~" QC , coi:~., _lc t:,' : = G k-,. =~ CY "s00 - a, 010

- . . .. PrCa a (2 ( to -.9 Ca

C_" CD _ w co 0 M (vO co w to M 'a MP' -1c 7et -" Cs O s- ICrv P* 0-me - - W~~ M. '0e 01(

o, -.-a I" .. -O

M* co o eo co C, C"c _ I* C ( C " Ca 0to - 0 la0 0 P9~'0 0C " -

cn~~ ~ ~~~~~ ~ ~~~~~~~~~~~~~ M tvl CL 3 C MDCZ * ^ _-. 0 C2 01 'M -n tfa 0 CD 6 _( c- c 01 0O 0D 0 0

-~~~~~~~~~C C6 ::I _r b_ C, to to C-_2. M (b C( C, C. 0 - -

~~~~~~M n wD 'LaC RD x UZ O X _ - V rD ei -- =x _ 7

" " O t ~~~~m 2 co cn rD co 0 --d~~~~C 0 0 .

a- Lb = ~0 ~ (0 'C -1 C,, C01~~~~~ Pg 01 01 0> s t0 ~ a- tj~~~ 01 (201 to t~~~ o0 0 a-

M D C O 4 a ~~' -. 8 P~~~~~~~~~o to 0 to ~~~~~~. 17 C'. I- CD

C a a M

. 0. Po t- co~~~C$OX_ -0 U D

~~_ b-U 1 (20(00 -* t-4 C C* *PC -Pgto

t -. Ca n 01 a.- .-g 0- >(2~~~ > t - 0-.X Pt ( - 01:

- ~ 0 D> 00,'.C -

~~.PI~U toW tfl Ca

0- .'_ Ps (2§ to' w: 0. m 0 to'.'Q _~ to* 4t2to to

W 0_. a ( .' P.Si' o pq0 _W~ aP~~t

-4C2-41

,4=1oPZ cs

1-,

0s-'0

MC=a ,-4

10co

0-

.40-2

I.4

c 0

to

m,,

co'gco

-u=

Page 117: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

I

e

SC- . c.-

~~~~~~~~~~~~~~~~~~~~~~~~''

~.~ ~ ~ t-. r~-, ~ .

co ~ ~ ~ ~ ~ ~ ., I'D CZ ~ ~ ~

-"'-rD C

0~~~~~~~~~~~~1

CD '~ D r

P-.~~~~~~~~~~~~~~~~~~~~~~~~C.k-.ICD DJ M. CD

c..,o c i- .

* ' CD

~.~~~~~~~~~~~ .. M = ="~ CD~~~~~~~~~~~~~D-¶ a-' '""-t t . 0

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~W-':: =3CD ~~ a V-4-CD CD t a ' -

¢'""1 a 0 ~..,,-'"1 a CCa ~ ~ ~ ~ ~ ~ ~ ~ ~ .~~~~~~r CDC

,.C~~~~~Oto CD DiDi CD CD

C- -4 M'CZ I- 0. C- i ' (-

o", --'. M CD~ CD '0 ~ CD CD o-.. a11M D cm CD-VCD D

., ,~ =.-=-~~~~~~~c, '1 -t- Q. -V CCD Ca '-1 CD 0M."0 C- 0 l C .-"M M M M ~CD a2D -V ~ aC

CDa- -- cog M~ co

I-'. C inn-'.'DJ ~ ' 0 0 I Al a, M - C- M M 0 C C CC. E. C, M MCD a Ca M M

IV to* C - C- 04 I.-=0 m = W. 'O a i i C

M.. C> 'IV a Q = i Ca 0 C-' I

0 " M M~~~~~~ ~0~' C- WI

CD2 cD co~~C, C-- - ID ! ,C -. a -aCD U~~~Di Cfl~~~~.4- CD Ca -~~~~-. =,-tC C-'

CD~~~~CDC

Ca ~~~~~~~~~~~~~~C

C - IV I .- ' D cm i C- IV - 0v- C

0 0 CD C~~D M- aD D ca -. D D 9 ~ 0

Ca * V~~~~'I 0.I C -. ' 0 -CD Ca, M a,

0- D Di W'Ca -C-2.~. C. 0 C"*, DIC. 0 , "-Ca 0 'o U0 '0 '-w CDM. 0"aC D C C=CD D 0 0.. C, .

CD a,. a~ "- - co Di M'0 - (~~~~ ~~2 a, C M' D ( f Mia C

o D40.0 '9 CD .. 4'0~~~~~~~c to 0

C M 00

CD ~ ~ , C-I- C-I- C-I- C-a- 0~ I Ca C 0 C-. - MaCa Di Ca 0- 0- Di 0-~~~ M-a 4 0- - Ca M

C-t' '0 ~0 M9 C -C 9 C -. a~C-C

C Ca '9 Di'0~" 0C M

gm.0 CaD M- C> CDM

Di (2 0. oo I- C r

-4 e-..

C"

CV2'""1

CV0("""=C11 C=I

Ct

eua--- I,-,-5

-- IU

C0mCV

Page 118: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

co Cl'.4 U

fD ~ ~ ~~ C C2coCO,1C

to 9 t'9 z¢2 p-I -C

0 ~ ~ ~ ~~~ CSa

o o

-~~~ ~~I _~~~~~~C IV I

C. C"booO~~~~~-Dm W 11f' °.

Cb. ~ ~I C''f-.. CS H0 0)~~~~~~~ > .C - CM .ta b O 2 C Ie .

M- M CC I

C* 00 0 C, 0 Mo I-co C12 '1 C O 0 II ~~~~~~~~~~n =

aI v * M.. D a " M 5 cor

-.9

O M " , t C· ~~~~~~ ~C ta DI cV"I - - cO ) CZ

I <~~~~~~~~~~C - ,co °0 oe M0 t

C, CO co CD -1 ' CCD

X Jvn~~~~~~~~as r s to~

Cr-

co C : O M C O .Ca 0~~~~~~~~~- r' a,0 C, C

10 Ca Co~~~~coC1

0 - C> CD co0

CO CO '.4'~~~~~~~~~~~~~~I Io

Page 119: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

I-d ~ , a et, I-9 -- C,Z .r c.m , coX C'2

a"I - lig 'm e1

Sa.~~~~~~~~m ,- - '(2~~~~~2

S~ ::s" c'~ C'2 t'~ CD co i I~ -C

- ,- C 0 la, ea0 COW a-CDCDCM.rWaCD

C-.-. to , ,0r aCO0 b Mb 00 a-2 CD C= C2rb = " - U U (2 .

'"' ~ -' aIav~aa- J0C "' I "0 0 I-' el to C", - -M, Eb a, "=I,--3 - -0 a-I a-0 ~~ ( C000 '-M~ a-,gm .

0C 0 a~ J -i¢-- m r o 0- 0 (2 c -

(2 a-a- (0 0 (0 6 ~ ' ~--, ?q' 0) 0 0 C C'A(0 U- -a-a- a-.~~~~~~=P- (2(0 40 00-0)0) tO a-- a-- C* ¶ -a c -

~.~,. rj~C 0 r1 (2 WO .0 ,t-a (2 0a -a f, O a-- ::0 -Q~~ ~"

CO 0- r-a-'-a 0 '--' MC - co (cn el~~~~C t. 0) .

CO '-I r1 -II a~~~M , c a n 0 aoQ

cn 0 C-,0 M ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ 0 ( (' ~ ~'"¢-'01 n rCO o a- 0- c

I-~~~~~~~~/ C> a, - ro W cn 10~~~( "I o c

M. coQ C a--, M C>0 -0 - eu

- " n C, --a - 0 E. a.0 I I-a-1C',

0CI>w '

- 0 e- In CD M 0 aCa- · A

M W-W C- CDa- 0 -2 a-a-.a-" (2 ~ 0) be~°~~~~~~~~~~~~~~~~~~~~~~~~0

1-- - C- C' C 0- - 0 : '-1 0- 0) D

- ~'1 0 a 0 m a-I ?D ,., a-I a- (~". 0 a - C- . C 1'--a0 '- -' CT 0)I'-m as WI C-a 0 V aV0- -. a--

o C -- o~ ao (0 to 0-0~ a-.0) O0 -%~~tD 03 a-I-a- 0 0( 0 C a-0) D, Co mccoi 0- 0no o 0- - - 0 a9aM- ( 0U O 00U ) ) 0) aa (aV w I C C0 - (0 0-a-a-a-a-0)~~~~~~~~~~0 0 a-v a-a CO 0-~~~~~~iiii

Page 120: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

CD~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~I9'O I-.I. co

', - p. C,,, m ,-~,

CD~~~~~~~~~~~~~~~~~Co C:to " 9LO

t"a - ' V..'- I

=, ~. -~ ,~~ -'- ,--0co ~ ~ ~ 'Ca ~ ~ ~ ~ ~~~Cto-

.o~~~~~~~~~~~~~~~~~~~~~~~C coCCa. ~ ~ ~ ~ ~ ~ ~ C ,-..0-,-- 0~ I I,-

y, CL..P-. ,- t --F. .-n toc--r 0 900

W .0

'" 0 MMC'-

,. - ,.- C-. -- .,,

m -~ m m n, ,-, ~ ~ ~ ~~ ~ ~ ~~~M o am -9.~tb c ., r.,,- ~'-'-~~~~ ~ 0 ¢'u m-~-- " m Cc, to ~ '.~ ~ ~ ~ ~ . ii 0 "c. . " -4 9 --d I,,,'- co-M, c c-- ,-', ,' ,--,

~~~~~~~~~~~~~~~~~~~~~~C ,'- '.-C Ca -

co- -CD q~

C-.-0

,..~ .. .D ._ ,-i. -C

,c ~ ~ ~ ~ r .40,l'l- c.O ..J~- I-.e, a 'V SW Wo 0 QJ

"'1 e d- ~' 0:: 0- . ~ -. I ~ c.

I"

M-* e~~ e~~ r-i.Ca rb Bi -' c-i.b t' CD M- 0'- 0 V n =C-i-

0 c- co. 0C-. 0DCa . Ca 0 -i.- Ca Ca0.

c- coaMC- co 01. I t-~~~~~~~~e,,, I=0 Q ::l, I -0 M - al ,C'*Y AP C:3 M~

i- I:a ~ :* --- - c - -I-- c c, . , C) M ., SU

M -" n n = cIv Wci M = M0 .' - -M r 2 C-0 ~ l" W cw M .)C.:0 CD. Ca cn~i 9C, o l=. SW tv 4~ - CD '-V

~~ 0 a.-(bC3 '1 A W C, Vn to Cz r- rti- co " n -r-ir=Ca '- Ca W c Uo z, c = C C- m cm. o W = '- (2 M ~ to .9- - 0o Qw m -tgs no =0 cr 11 Ca 0 ci. c-i Ca

c- to 0o I.- 0to C a, CDC3 1

co -% Q - I oI Ca

Page 121: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

I111I

-' Z a,CO9"-, - -

,-, 0- C - 9'O :_CD _--Co C C

C d-_. CI- 1cr' 1-feO C C

Ca 3 O

c,

.3 0.~I 0 14 0 0 0 cm. CDco CIU 03 C W S= co co coCa Ca C, 03 0 I-= Co -:i 0 = CD

l 0 0- C 0. .1 0 C O DuCDCa co (O 00 C1 D C-

- C- CI * c,-la .-C C- t.r O

0' t~ - -- V Co D '" 0I:0- C-I ¢ D . 1, ,-- ea 0- CD t "-"C

CD 0U MC-C Mw- scO t-9' r q k-, Pt--'f0- -o 0o -- C

03 I- C 0 a. 0. 030 ct$ b 0 C-I 0 Ca C Di 9-I CL b Y m X X ,_.~~C C3 - c

-- 0:Df O" 0 = ' 9- - lL3-9- ¢ - '-

tn~ I~ cr eS :~ o_" D~*'19 ~ ~c-9 E J .0 ¢"- 'o O C1I0 0 o D- 0030303 _ _- b RC, C' I0-t~ 0 S ~ 03t ~~~ - 03 ~ O C-I03 0- 0 > 0-V Ds - p C 03 I=C-I-CD .-CD CL 03 M C- -a C- Ca

w M _. C> ;a -b Z. M M, C:3 - I 0- _-~ 0 . Ca CD 0 -IM CD CO D _

C, CI- o- 03 to C C C =a=

O C CD CD M D C

t"~~~~~~~~D =v 'AO C', >C D M * D CDCa 0°a C-I- - ,c C> C-I-

C-C. Ca -0 O 0 * C

C -g0 0r r- . 3U0 a C a 4 o 0 '_9-1 C-I- 0- 03 C C D

00 ~~~ C-I--0 tCa C12CD - O - C C-I- -0 ~CD cnW 0b

D _* O *_ ~~~- I !: fDt t- O :t X D C- 2 DI CD - 0. C-I-C-I-C-I-

D b C tE O 1R1~~~~'b 0- -. 0 CD Cfl 91 . 'mCCa C-- 0- -003- 0 * 0.C CD(D C-I-- * C ~ O

tO D 9- 0- CD O0. 9- D O

Ca O

0 0 C ~-4 O0 0 C0 Ca - C~ 0

C-I Ca 03 0 '-

CvD l C-I C09- _- 0 D 0 b

0D O -CIC _ =ro )

. C: D 0 0

CD 3 Fl- 0

Oa CD 0 "I-" 1-. O . C

'. 1- C--C

~~~~~~~~~~-I 0 0. Oa '1 _I b . D t't> X 0 0 (CD 0

I~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~CI

C, C.CD'

0-a0 C'

ao a03 C0 cO

0.!-. 0c S_

e-1.0Cn n

- CaC, 0

CD_-I

to MC Ca

r

09--

- 00 CD

ac>- M

.- CD

oCD ca

. o0

C-ICD a

CDD.' 0

- --a

M. D- C0 0 0

-qI - -

-. 03 CDC2 -0--_ -.

C. 0- C- co

10

co

.~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~~.C. C4l M C-,-C s 0 -*- --

- C - O 0

cnor coW

9- I-i C-I Ca"C-0-000' * 0

0 Ca .03--CD

Fl_- 0'

CD -I C- I 0 0D U D

0 II c v 03 -

0 0 n H CD

0 -9 co co

to o i a

* -S a* Me M n

C-I-" co3" 0 0 0CD D 0 C

am0Q cm c

PI to t 0 C

MC -- 0 0 - _-t 0a '. a -* 9-CD - -a n

0z CD '. 0 0

03 C : O Ca_

C D _ _.3

CD D -1 00 CD 0 'V 00D- 03 Ca 0 03

D3 0.0 O CD

CD C, D C-I-I- C-I- Ca

. s 0 rl- ~ to

* CD <9-1

Ca CV 0 < -g

D S: CD _. Ca * I_~

03 - - C, -

03s D C-- C

0 C 0

c, '. tD I- 'V

D~ C-I OD a0_DCa - C

0 Ca 01-

00 ~.S~ ~ ~ 0 _ _C_

TC"-MO c- C-0 CD _oM, CD

n 0.C a -.0 0

Ca C.C- A.

(D 0

D~C

O

O

C .

VC-L"

I -g_ PO0-4 W4

C= -D. ..

N..

__ _ L_ _

_ __ _ __ _ ,

Page 122: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

~~~~ ~~~~~~~' - ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ .6 C - (1s,-* r. :. : :

a , W P "1tv I~ "fl' 0 e"'- 0"

0o e-I 0 ,- nn ..f

00 a) Ba C - , , =6St3-- :~-,- ~

0 ~~~~~~~~~~~-

W- co

c'q'~~~~~~~~~~~~~c

M" eu (,9M Ecoa,- CO4 C~~0 ato 0 co to f'.. 0

-. ' IC)- -. t " .o.- 0-g .-., 0 -- 0c -9o rI- C- -CD ,. W t co

me= M c 00 f--h c* = 0- I

to t, , ,,, , ~ o -:

:-- P.K. rl = " t -co CP, M ,a "aC' C, toCD C) " - a =C I~- "

I'D M "2 C,) t,,-e C..~ Ilm Im.. 4 C '. * 0 . .-. ( 0.'03(M

Mc O .n 0 CJ - -

:" .. ,-. : 0I C co-'CD C. 0co M -0 0 0

ex,--, k.* . r '- i- - - 0

~ i0 W L'"f0-- I"-

-

-C0 lo ,.. -1(J2P.~~~a) 00) ev(2 C. W f--a W 12UUM

000 00 '. IW0 M0 ca 00o ..0 f-,- ~ 00 0 I~ '-,..0 (2 0"0 I-,. :-"'1 C.- to

h-,.,,, - "10 * ~ ~-,.,-0 , '-r,- 0-

M 00-0 I-V -. I- M 0D -if-I- 0 00 f-I- 0 -~~~~~~ 0' 0 rv'C12 C- -C) c f-- d

Cd2U 0 '---0 a'I~~~-- 1' ~o -o, I i-00 Ba 0 a-- .1 r o = 0-)0 0 =~ I -. B co

'~~ OP--f-I-~~~~U coI 00W IIC>CC*

C -ft M0 . E '-( CA .-.- 0 co W0 4 a

W eb C(-'-(2 el -0 OP---M c

f-CD, W a ~ " -- B V, W I-V '

00 I CJ~~~~~'-V 00-~~c M 2 ccm- _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 1. to_ _ _ _ __Mto_

C a) '.0~~C co 00 0-0 V 0

U a- 1,. -. U B 0 .- 0 aI

a, , CD 0 g -- 0 0 -. co -I =) "- -

oo M C2 . I- tv W0:

Page 123: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

Ill

.. . C.

(2 r~~~~~~~~~~ CD a~C~ o 4.C D C CD mCD -. ~~~~~~~~~~ ri -i~~~~~~C00~~~~~~~~~~

Ca CD 0 II~~~~C" C C: a C,

0 ~ ~ ~ ~ ~ = ~.w 0 ,~-. ,-,Ca ~~~~~~~~~~~~~ ~ ~ ~ ~ ~ ~ ~ JC - 0:

o,-0 C, Al ~-C O

-.,-~~~~, o- co , .-=

~-) 0 ~~ C. ~-,-,- ~II~~~~~~~~~~~~~~~~~~~~~~~~-ICD6 0 " CD (2 0 -a 0) 0 CD .

I:0 (DO 00 ,-ii r- '.. . '.-.

'I' C-I" ~'~ 0 .- (2- 1 a-s - 0 (D0 0 CD C-- o ,0-b -o - c.... ,,- ) "I I'I h-C0 * CD 0 "I C-co a, C

2 rC C,-C (2C C "20 (D (

CD C '3.- C. rlI :m IIC ~ . 0 C - D CD 0LIV C, coPft) 0- 0 to0

oI i-,. I-9- L . r-M = ~ ~ ~ 0 - *- * coI0 C',CD IM C ~ ~ ~ CD - '" 'I 2C.0 C

.C0(D CDco (

,-I ~ k ~ ~ ~ C-- CD 0-- ml 0' ~'

CD M C-D~ C 0 D~00 = b- 0 I Cl C_ _ _ _ _ _ _ _ _ _ _ 0- I~~~~~~~~~~~~~~~. ;-C01~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~1w rbCD 0 C Q " a-1 .- C- i II

C Ca0-0 C 0 Ca0( ab MD01

E~~~~~~~~~~~~~"C o c (C. 0, 0)M

--- C0- 0 C-.:. 0 - C.

CD 27 e--C ~ ~0" C-C b-9- -1 0"-. )- .0... 4'"0 CD a -IW- C0)C, 0C

C. C-I n C.. 0 a=,~~~CI - c-D,

0) 0" - M - b1 1

· -. - · . 0 CD

0CIO I Ca

C>00t 0 ° C Ca C.C-I" -. C-I- 0- 0) o _ C 'o - U

w C C e' C M 0 -~~~~ CaI Ca M--C ,-4C C",a

0 0"0 (2~~~~~~~C 00 = 0Ca Ca 0 (2000, CDC I c

* 0 0-(2 * C-I" 0 -i~~~~~~~0 CD (2 CD DII C

n ll0 0c-, -' 0CD 0 Mn2 =i (2 Ca (("2 co CD -4 M

C--C 0) '1 Ca C CD- -CD~' ;-" C-I - CD 0)v CD

0. CD0 C0CD D "20 0C to n. f=~ 2 0)- a- a i C0 -- 0 I- CD - 0-~~~~C C C>C -I0 CII

Ii 0 0 i -~~~~o - = = CCD 0-'i ~~~~~~~9I"CfMlD W -00 -

0) CD CD~~t~r CD CDC - I- =

C0 bi 0a C> C>9 Ca -

co.b- M C-IIo. CDC M c-I C-I"

CDt C- c C.

CD D C c CDD C 0 - D 0 bl0 C',0 i ~ b-I 0 ."4 -04 2 b~(2 - ~h ~ C

--

-4

2

t3

I ¢o1

b-II0 -

'I

"'

-0

C:Im

Page 124: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

~~~~~~~~~~~~~- tr'f'

- 11 0 .0f 0 0-,1" _. P: (" P,(2 o r CD 01 C

o a30* , -. . -.

to to Ca a, r" CA nM w " -

CD. , -o - . , C

ce) - ,--. U cP 0 - CA 0-,0D A' C= --

:3 ¢"" 3' ' 1 C k.- ' I 0- . i..3- (*0-. CD e- -t-~ I-4 1B a -. 0-r3' ( 'C -o C I c~~~~~- O ~I--4 ,

e _ -*C c -- o- cD B C CA CA -B 0-C B _* C t

C (-3' I ~C _(-- ',- -. = . c0 -" 'co rl o w: . " M "o *to _ - (D M 6 co * r

3-v " C C C_ * 0 D ' 0_- C D co M C. ce 'm - CZ_ CD 0 = - -0D C r CD Cos _- A -I

O1W c CA co WDC 22D- 02 0P~Ca, c 2'2O- - -33- '0 C "

b43-0 _ .- 0- t0 I.0 COo AC CD CIc'- r3' - *t 0- B 0 ("20-( C&. A O et,co P¶M -. . ,C Orb o 0 0 _ CA C P¶ a o 00 3 <n co O n> _: _Ff C 1

O .- -. O C - SWC 0 _ C I -c vOOt -bXo ICCo - -C _ 40 - " " , M _ ' ef _CD ( CD' C 3'-2 . et 00 0-.-co

~~~~OCtD 0 C t v lt eo(D"nO C CI ("2_ o b c0 CZ CA0 C-

_- - o- - o

I c

C0D C CD C 0 _

Iv D~ *D XcD 0 * v -,> ~~~~~~~~~~~~~~a

OI,.. .' CA O4 0f " C" I B _ -I'1~ CD r - o - I.-0) C 3-D _

-_ 2C - I :

Do I C100 I _ - 1' CW01 CA 0 o CA . s _ _. 3 CA - O-4 CA 11

l~~~~~~~~~~~~- _*O0 CD ~C ("1- (-D I '00 0 C4 0 (beDn - WI

.0 ( CA (-3C B:e 0X D

!~~~~~~~~~~-- o-- -va (-3-

II ~~ 0 .- '. CD0 CA~~~~~~~~~~~~bC _~~~~~~~~~~~~~~~~~~~~~~~~~~e no _ C121l~~~~~~~~~~~~~~~~~~~~~~~~~~~. c , o .. M

0 D , m E1DcC. Au co

01 C B Ca MCD -'~~~~~~~~~~~~~~~~~ CD ~~~M$ oC

_ _ _ _ _ _ _ _ _ _ __ ~llM 3 C I II~~~~~~~~~~~~~~~~~~~~~~~~~P DI0- to

Page 125: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

ill

a--M. 0 a-"ra - a--a

a-- = a, 0

0 -- '0-jma( 0 aa

=0 a- ~ ~ = = .~- a0 ,0 - .- oa 9-~ ~ - n M r c r 9 -- .- 'E . S Wr W' ~ ,,

~:~ f~~~C, ::~ 1 CA 0~~~~~~~~~* ~ ~ ~d ,.,i a, 0- 0, a-- a a-- c ,d ra , 0-C &~-Ca a P1Ca C~~~~~ Co et-- a 0-a f C aSa , - -. I j ( ca a. I -i a-4 O . . c 'CZ0 a .. C=0 0 0 0 ¢ ' c-a.a, to-*0- a a, ra '0 0 a Ca C

Oa a '- ( 0 Ca a a' a ' CZ a -'"-a, a a~~~~~-i- a' ¢' = - .- ,- ca a aa'-- a r waaD 0 ra -~ ra 0 0 a 0 0 a-. .. C.C a

~ ~~~~~~~~~~~~~~~~~xto- w D e a 0 . C->i 0 ai U , r -

a : ~a, ~ -. a-- , co. -. =i0 u - 0. C C, 0 C cIn 0 0 a-~

co Ca ~ -' - aC 4a a a -toC 0 WE 0-. CL . S. a a, (a - -i

0 a' - = " a a-a a -- a- W. ao ,ton - c-a 0 I ae a--aaa 0-0 C, t- C- C0 ra ra ao~~C co-a m C", - aa a, 0 r

co-ia '-10 I-C to Mo aMc'- CD ra'0 a, '-ira~~~0 a-D 0 am CD I - r- cn I

co ~ ~~~~~- CD ~1 D. gCa I '10~~~~~~~~~~Irba

Ca i-V -I a C 0 a a--- a'i 0 .. C > -0> ra ra 0 a C-ac CDr ra 0 -ita"

R-i A a m raa a-r rao- =roo a ra, -4. - - D C 0 0 o

ra e0- C-0 0> 12 0

0 a - -6 a

ra Ca 0 Ca U Ca Ca Ca C- C raCD a- M - co 0 a- (a co.

ra ra a-. a-i --C a* 0e - C

a. -1 raco- a- F' a.C

a-i 0 0- 'i'0 ~ ~0a C Dc 0-0 r a- 0 co--0- Fa r .

C 0 0 CaFi-a"0 F

0-a- l -i 0 aao CZ 0 0- Ca 00. Ca -* ra a, ra a a, ai - --

*O a- CD-I'- 0 -- o a co

co 0a" - 0- ra a, 0 C0. -i .C> = -0 0 CD 0 Z0- 1

-4~~~~~~~~~~~aia, a~~ Ca a, Ca -* o~~~a 0 m a-* 0 aC , -

C, a- - n C0D - = a -* ai- a -0 - r = c o a-a- CZ ra =

a, Ca (a a -a a, '0 -. ( -.' - , C CD a- a * r a , - 0 a

ra r co ~ co C 0- C 0 Cra wCD-0 0 toa ,' ra -aI

r 0 >i 0a, 0 - Car 0-. 0- Ca 0~~~ a, a-* 0~ a et- Ca , -0 ra

ra 0 0 -~~ ra C0 0 a - 0

ra Ca a, a- 0~~C C:-w~ a - 0 m0

0 ra a a-o 0-00 -- ra 00~c Ca

a, Ca ~~~~ -*~ 0- c( a- C= a ra 0'

Ca Ca Ca 0 '-ura~~C CDa ~o -

a, c-a- (a t0 coa- C a-i * Ca ra a-* a- ra C12~~~

9:. a- r"

0 - C a -,

~- a, r'0 a--

C- a,

a-.C,0to

c. , c. co a, ca. -o

raD-6 'a Ca (a , 0-. 0 - r

co ~ ~ ~ ~ ~a --- 0Cr

~~~~~~~~~~~~~~~~~- , -=* r '-0 C -" Ca Ca-,"- 0 '"1 g~ - ' 0"

0.--- co M , ---. m 0 0L -a a-, r -

r~~ CD 0 -rt, ~ a-- a a,

a-·0 ' 0~, ,-, '

cn t3 tm C - a- a t=

I- n arr

E .b· ~"- ' I ~ ~ ~ ~ O

a a r-b. a'D a c'a--a D 0 -I a, 0 a-a- '.ca w wCa0-ara r

- .-. ca atoa Ca C2a C

~ 0 a-i 0- a-a- -w -i a--0

~.0 a, 0 0- ~w

0 0 0- a-i a-i 0 c-a- a-~~6...-i

r, aO --I a ~ 0- - c'-

'0 ra a, a--~a a ra u C a-a.

-w a, 0- a- 0 -3~D a-

ra a--Oa.0 0 - 0 Ca Ca-

Ca to0

'0a-.,

t0

c'

-i

e-.

C,-

tomB C

m~o

Page 126: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

C I ce a O c o cc -ea sxIa. a " S c t O tb I :S: cI 3 a 3 P,3Co I ,'.,,'o>C3 =, n, o .o ,, ,

C a _ -" O D _ t D -9 - w M *

3 cX X _ - 1 a 1 _-C tM _*· mp XC1, cc a r. I ao a~~~~ r"-, 3 n, , CL. n -4 CD -0 0 "Z CZ S2 CL a,i C - " C aW C7 C . O

. i b C .- -, -f r .C'O c" t_CD D t p C- 1 , - C P- 0 o

am O MO O to CD CM D

MuI a C,. C- o C. M,-

I:~ ~.~~~~~~~~~~~.. C7c. I -I 1..d.~.,~l

CD t ~1 C e*_* X Cc f ct r ir _I * D O I) I ! 3 cc PI CI

r· O i Cb iv a P> 3 c c. C) CA

" a 92 . - C C0 . Co 1 Co to X CZ.

. _ O -= t l -I O V-· I i; S 0 O cI a a _ ca -

C, 0 i l O - a, cn fl t e C w Vc -

c v. I _ t c 0 o. .m = C w C aIf~ : -(ed ( 1 _ ~ o-,- :::s ' C I tC C ::Z't

o r * < IS , ct _- p, a 1 _ t i s °tco rI - I cr a t17 . --o c C: Pt or CB ccr o ce Ir iv a c - 1 E * - : I M _ ro M,.- ' -4, Q,,C t ~l , M.. CZ w co .V " _. a c M C = s _c~ I,,,, · P C a, ", C= f w CD Mc

~ I _ _--. r· _ _. i * ° o Cb I P a a c·)m CD cc a i Co ~rCO'D

_-- C es c _ _* r : ·eC a uh X pIa> a Oa -* a 3 C _- r I cr a |c 1 a O a i r c- iC O3 i t Ic

to _ I D I w S.~~ -. 11.. o oC~~~~~~~~~- C Co to b " :w "aa A- C

cc >o _* :r C O a a c 3o I o o o ·@ P, _c > _g cO o O CI n cr _ 3 cc on IcD a o CD I a _ _- c :r I a to "l cco X s 9m " i

3r00 b eV to

X cm X 3 I X I rr, cc CD w reD~~~~i a I o ov iV 3 ctD I o l-o I 3 rr, = = M CDto C

w Co C$ = _ , e o Ve

'~~ oD I g ° to pv g jm cc I Cs a C: toet

C'e'~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~~~~~~~~~~C

r~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~~~2C) C= C, Cb 1 o, " =r a.mn - n M CD.,Co C= Co Co a, = 1o M UZ to C.cc b I rEoWS e~~~~~~~~~o_-O~~~. ootccm,(I C 0 c CP C- Ca V . C-c " _ 1 cca

w Co nc CD

C2. C, - to to

a ·-- Cb ~r I w -b O C = C D 1C,, CDo tlf 0. _- s _ a fb t D CDC C3 I C.

C CO sb =X ctt C_2 _ nt C" ,a 'mr o n I& bCD _ - r o 1 na I

or CD to CDD a _ Co iV *_. , . . _. a cc to I

O a C a -~ C vC O Cr_. I=$ o C ot D3 ( e sw I Pa o C,

C -D Mo ro _ 1 c a co ct r

C - _.C- o _r ,_. a 3 0 - O CO I C 1 C abf Z tO I m iV In

M t r_ _ I CD a a I

I I <_ _ O-eo I aC CD , > Ib _- CT _ -t ' D a X I XLt o 1S 0V C.

Pr I cc gc .o r· ( o c 3 > b C~a I c- cta I Co I Io I t o

I .- m , _. _

CC C o

_3t oeb § I 3 I eb I

11 O C eD C D

11 CXP I ' _ D I oo I CD cc cr t a~~~~~C

to ° I 11tt | D l roo

11 br I r 111~~ ~ CN I 1 11..~~c .· . . 'I

9 "W

M

to0-4

P Deus -

v

to -90-0 an

IC '

Page 127: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

III

-w0 - 0 'I U f C 0i 30 a 0 , w.0 cc 0- 0a y ft -c 0 0 0- C, 0- a ft~~~ a ~~~~~~ cc I ~~~~~~~~CD Q CD I o a c g CD

cr 0 .4 _ 00o a ct I -f P 4~~ mP, o ela. ~ ¢'0 .-1 a 0 e' - ft 0-0. ( a' .C. Po0 t CD to a C a

=0 I

l .0 c1 f a C ca Ca " 0 C" - 40,f f '. CD C " 0 c m

4= o O O E C. to L

a C0 ft ft - 0 - C. C 40c

Co , C> Ca 0 o~= -- I -rrl~gOI a ft e -

.. - : ::" C'1

co a 0 0 a cP 0 0 tio0 .0 ¢2 ¢'Dt CD -S.0 fc ft a, r- -ica - cr 0 t ft 0 Ca 0- (ea -a o - . a .,-rP C- *O 0-C Ca

'-4 CD c cP % b 0 P4 ft 0- C-PaD

'.0~~~0 C·I ~ ~~-P 0 C"I 0'- C D :

-4-0

.1~'<~-¢.-S M M~~~ "

~~ ~

a~. -'t * CD a a

C C , -I ~~o0 0 IP \.0 (2 0- #- 0-' m C-' Cit ~c '-4 -. as 0 Ca CaL P1C a'0

0- 0 0~~~~~~ CaQ =- C -C- -, .- -,.-. ~-~c~ . c- -a ,- w.,--.a' 0~ 0 -0 IP a~ c

CD U 0 0(2 CD 0II e, cr I L·L~~ C~l p)L D, ( cPa, M C1 r· Ca 0 (-4 arcD 0 ( 0t' '0 a' - , 0..C-· C a' 0 r.

00IIc to ca u S. ,.M 0 I *c ft P1 ft a as Ca a' I~ :: -0 0-cc ( a 0 t r0 -.*.P 0a Pt I - 0- 0 ft a -* -. 'm~ ~ ~ ~ ~ ~~~~~n C>cc o (0U 0t U t -'0c 0 a IgrO mp~~~~~ , a M c Mm~ M C a ' -0 ocm

co - 0 = =ac (- 0 a (0 a ca

rdo I c t o a 0" R' 0 as a a, 0 M '"ft0 0, ~0-0-a a a a' ·'0 0

'-0 00 0 P

0 . 0- C Oa M U, t " =0 I CD 2 n '- C. o , a M .f 0 0- -- a' 0 U r4 0 - r 0

C., O I- I - C. ft P4 aM a It a 0 o a' o 0Cr e- w-n ft 0 Ca CJ C2 ft . , as( C (- )f0- c-· ) a s 0 0

a, C 0 CEn C, CD) M t r ft C > ft 0- ' * 0-- a 0- C) 0 -ft cn a' 0- 0 Cfl fl a'ct as ft CP ft' n a ~n -C 0, f 0 a Cs a' aC-it (- -0--C-' -' 3 NCc aM 0C- -a C M C C '0w * Ps Cft 1- cs ft ft .- 0 0o c 0 ft cs2 ,- c-5 r* U 'w ft ftM, U2 -a a, a, I f' ft 0 el t 0- a, -0- 0 M-0a' C> '5 a a a 0 - ft (4 ( 0 0 a' a: ft -; C-' C 0 0 ft

.Cft .. ft -C,. f C f M, 'A v 0 I . P4 a ( ICI r 45 r 4* c~cr I a -C c C 0 (- 1 151. 0 * a' 1.0 f t - ,c a' a'

1- 1 P5 C 0 I- - C, ft = r 0 0 to-~ ft "0'.0 (2 U I -n0 - as, oa IV f 0 0 t 0

1.I L7c -· CJ¶ 1-..) I * Ca ~ ft 0 0 0 0- 1- C-' 0- Ia U 0'cc (3' o, I-0 0 0 a a ft 0 0t *0~-

0 D IC C a P M Ca

v, ~ ~~~~~~~~ 0c~ ~ as 0t 0- -4 '-a 0- 0 C C) (-4 0- 0 I- -3 QD W 1 0 C C2 a, 0 C f a-

Ca (s c~ w er I 0 a P, 0 -* ca C) U Cc C)1 .a cm Ca . -C7- l f _., CD a i(' _ e- coM

a, r ri- o a ww 0- a' 0, ,-- C -a,.ii~oo 0- 00 -- f 3 U0 o o 13 o o o o 1* P4 3 C -. I 0 (4 ft ft ft 0~a

cc P, P, ID ctft S '--a' Ca 0a' ft ) Ca 0 93 ft-0 I ftC I C

___________I _ cin CA. C, O =a I0 U· U a f l'0, C. a 0I 3 ( P 0 a r I cr- -. 1 (-t as ft 0 -C 0 3 I CD C

-~ ·0- ftC- 0· ft ft- r tCN~ ~ 0 a O0C C~ P-P aP P c ftc IP

a 0 0- '0 c~ a f 0 0 Ca'a 0 0 o co0 ft 0 U~~~~~~~= ~o(2 P4 (2 a', Q Q E as -.

ft f0 00 as w 0 =W 0 0*--- 1. 0 ·P-* C 0 ft frr t * -0 I0 Ca 0.0a- a' 0 -a 0~ C3 '0P 0 Cf 0-0 1- U a'aa a r·'.0'0 0 0 a ri cs2 - v-- Ca ft ' CD w M.to- 10- ft -. C2 as ft * -U 00-a a' -. -' 0r 0 0D . = 0 93 Mw oC 0 CD

00 '0I ft 1 0 (1- Ca 0 ft as Coa aCD ft0 c -C ( a" ft 1- (-4-'0 Ca (- 0-

I CP4~ i 0- ft 0 0 ft -- ft O 0- ftc Ca ft 0 -* ft * (2 0- 0 0 a- 0 'S a' pM P, U

-c as 0 ( Ca ft P5 (*0 Uc Of (a C3 c ,L cas (-4- Ca a' uPr P1 0 0 '0 a' 0- a Ca Ca I· -

* * PS ft ft 0- o-Ca Ca Z. 0 CaI ~ ~ ~ C c-p ft P) -i -0 0 (--a -* 0 '1C, -.. a C ft d- 0 0 0 a'

C4 a IV 0 0 C .- a 0- ft M0 - C- - - ft f '- as 0 ". .

M 0cm0- ft a~~~~~~~~~~~~~~~~~~C~c

Ca

-g L- -~

-1-.M Ma. - v

!%a-,

M-aM to

CO 3

aM

M C)

M:L-

0POLPO

0

e-to 2~

Caar

CI,:v

Page 128: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

'-a 0 b-.I 0 '- U 0 t' 0 0 a ·0 ta -ao -W ' 0 , , -0 o , - a a , 0

CD. g=0 CD 10

C>~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~C9(20 = M a a W. - 0 - =O... a-i-o a~ -c0 0 0.,. 0 =a 0a' n II CZ 0 0 = I a a 0 to ,, 0 a. 0 W--- C- W-b ID"a

I I~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~i-to tW to la M I b =. aCco um r" go co .- 3 0 'a 0 0-

lu ~ "e S:= ~ 0~~~~~~~~~~M. -01~ a 3 00 a "

a-a S:L 0' e q'm 0o II I a-a '"-1 a'C M S 0.cI a o crto M

= -'a , -a ,a -a' 0 '-to s -i ( C

a- -.- a I. -.

~~~~~~~~~~~--0 0 ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~~~~2C 0 0 '"0 a-a - 1

s-s a 0 3 ato 0 a- ,

~'

' 0ao ucrc 0- I ____ I r I1r 1C,-CD, ;' '- Ur °="

a-i a- 'r 0. a' a' 0l a-.. o. a G - c

~~~~~~~~~

o. = 0X C"9' aJ a-a- * ' 0 a-' I -"I* f" 1 0 a .- a' 00- r s-i - a a' a' a-a ..-. 0o .oato 0 i a n o

-. a'b-.* ' t 0 t b-a'.0 a.1 ' C/)a' f'l gL, a-I- I..-4.a I '' -I

0 . -- ' a-- a.-. ,.. .- 0 to r . ' ":~"""r I .'-,

(' I.. -'9 a ' '--a

CO a' a a' a Ic v pr ~ a a' . a a I ' a 'as- "a aIa- 03 Ca' '-a 001 O I31 a ~ c · P r 0 iAPn 0 0 im, Ca' co a o -u t ouW - 0.0C-

oa o .- =-.a a' - a a 0 -* 0-

, O t0 a -* 0 -. 0 a' O a V'a =-- a I 0P 0- 0 a-* a-5 a' a-a a-*I 0 0 a' 0 c IPO P I C*C fa a a- a-a-so I ~

alC, - - .- 0 =-- ' a . 0 t I, co =

-. 0 -~s a-6I- a o.. a- -I 0 s-*)3 ~ L It ~lcc cA 3 (3 c~O ~ - c U C aP C.3 0

la C> 0- 0 W W W a - a' 0 - -. I a M W' -4LLa- 0 o v-.a' a (Da

"'' -= a' 0 '-

'' °· °=' -· ~ ~ e c

I-.b s-' 0 '-a a a' 1-'4"d 1 0 0 I t r 3 1r·C e (? Det, (a- a- 'e- '- 'M.M coa 0a ' C C2 = 1 U- a (-'-0 s- a. t C, C InIt

0 'l C) c " t~"* W Cft,M, S eVn fC~a Pr · 3c ( -cc N m c3 I 3 PI Pt a W C2c r co t kd . ei. W a' C- -V 0 .. f I1r·- o 33 c co W to to 3 1 I a I

.-. _c 0.. W

a' 0~~~~~~~~~~~~~~~~~C -

c cia a-a mO s-a '- - a + Q a-- 0. 0 a- -0 ~ P 00 S t c't '0a s-D a' 000o 00 0 0 0r I' 0 0 -a5rur --- a a-i 5- 0 '4

a -']- C: CD k-P a-Ia' IC 0 II a - - C I ' CtIC-a' 0 C- atoa- va -' 0 a

0 ~ 1 cc -' - 'a'00a-a-'. a 00 coI' a ~ 0 a I a- 0 0- 0 -6 a- 0

'4t 0 o co a- C 0. C, a' 0 0 0 ai '

W 0 aI a· 0 0M ,C D' -0 Dc a. Ca a- 0 a' a--

Ur -m o a 0 a--s '-i 0c a-a- o. I I e c

- r~ a- s ~ a -J· a a- r~ia 5. a- 0 0. a- a--a'~ c a r·Pr 3 ccI co =c a 0 0 0 s 0. C- a a a a W -

v .a '- 1 ' 0a- 01~ r oa' W -a 0. '0 -a a

a' -- so * ur Pco C I3Pr = c

co a, co -- I a to C - C .

'-~~~~~~~~~~~~~~~~~~~~~~~~ 0C. I c ccc co MMr to Mc

O ocrn3crO acci rpco r3 r O I

ci C cm C 3 O C r0 W c CL c · CM0 C6 X C" I :4 W W a. cor O t P I D ~t n, a 1 C., CC C CIaa13 P

3 cT ~ zr a f- eV 3 t( I 'M W CD co*c ec-r-I -· wc co O = t t m 'a Cm.. owwtO C C. . 'a a cc W. C, 2cc I cto cl 9.ft te o -s 0( W r· I

co P, re ~~~O O c~~rO ~ a I h, Pr 3 SW co Cc Pr QD 3 O Pr · L cc CD I- Cj C- ?la r-,

P r- _ _ _ _ _ _ _ _ _ _ _ Icm 1 r-- %V, - = %=Su c o C)- c to r a -

e- CI 9, O M I co IC C- C, a- 0

p, PI(3W~ ft :v,3(P ca to -C(2a o Tr CD C. C- rr CDII ~ c , c- ( r O -v 0C, M Ncc0 I co accncc r

M Q . Ca c L. a a3 W WC- wa

a r (P OI Cc TI I a pt C) a, -- - --- Co 5r· · crr 3 o Pr I , DD ~~~o SW c·

co -r4 r -7I- W cv M rOI

I p 3 co I·I C· a a

0-~~~~~~~~~~~~~~-_ _ _ _ _ _ _ CDMI __ _ _ _ _ _ _ _ _ _ _ _

Page 129: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

III

rb . " .I M -- 1c-,0 PI CD -g %= CD -0~ a m~rZ" ' D r- a 0 , i C0 l C-04 Si o ctCD -4 CZ W = CD

C ~ CD 0 CI '-w 0 c. * ~ Di ~-r- ct a cC CC'D e- I CD Ct C CD - c CDco q ( 0 - co -- n M -wC)

to C".~~( MtC "- Di M PL (s-= CD ~to _ 0D r, -_,

.. CD M - ' i r CD "a -· 0 .CDD' CD. C>D = r

_ (1 .- ~s C' i s-'I CD sI~ e . - ~.. CD . - c-e- 0--Ue, I (

Di r S-. c 0 CD s-CD CD 0.. s-.O CD irs-C. 0 ICD U ' ii~ -C CD C D ti0. I

CD M C s-C - 0 C -,5- M MD Di5o M 0me 4 ' *0crru, i I-. U s C C D i C CD i c ) Mc C" C. a

s- ii 3C CD b4C - C CD t0 - co ~r Dcs- 0- i I nCD Di cu n, (s-acD s- -a r-s- CD Din c

CD c n cD ii' co CD

U s- C M - -C- - (2- .i 0 Cw -- CA. C' w i 0D CA C M c o -M IwcmCI- W n C:- C.-- " un '~' 9 ca . 0 w M

Di ~~~~~~~- c to C" C v0A CD= c

-C',. CD C ('M C'2 0. CD :-CDC CZ I iCD 00 i SC'-rs i- C "v1

p, ct I IO Eu O O, a a ct C~ OI I t' r Lii CD (2 M CD CD CD 'D

CD 5-'. C CD-. ( - 0 -. n <.=a ru · ( D a 3 u r· , eu (D n O r ) C-

0c a(2 U ( C) 1 tv Dc Lc CD CD D

co CD U co 0 8 CD tv ( CD U CD i C

'-- ~ I~ ., -" 1 -.- O,

MO -C r, v M r· c-· I" - I co M - VW -

° =, c-~ II·c- c -· 4 r tOnac C I= D sq ( -CD .CD 4 C C Ie 0.CD -cc ~ ~ - cr 0. o cP Di Cc r-i Di CD cc50 m CD C C D t1l D r Di'CM., CZ a r· ag c CO · u c

Co c en C- CD - 0 s-v CD I D =r-C 5Cs- C= s- - C 5- I CD c o (D 0

M D w Vs- CDs w-C C

CD CD Di 8-. 0 U CD ~~5 i-isC- 0o CDL 5' . -.

s-- s- 0 M- ,- -0 s- C-p-, . SW llV0.0 .- rs-- ( aa P, C~ C c-i- 0. _c w w rD `4

CD U Di - 0. S- -~~~~~~~. e- m 0 CD>D*o 5e cc i c-· P, a cD a a 1 I =r CDt= CD UC D M S .

s-i CD~~~~~~~~~~~~~~ C, -a CD i i rC3 s-i. P 0 s C M CD - '. CD 0 _ ai (2 w-- s-i Pc o D -. = .* - I3 ~~~~~~o ct I C a 3 a P~~~~~~~~~~~~~~~~~o m c-- , M wc

M CD n CD D CD 0 , 0 CDs- I ' S -W -D s-~W I U O -0 C D = M C 0.o -. - n - ,- a --C Di 0 D C o cCD * Di 00 Di s- aC i CD -- 0· (2 a (i Di CD Di U ct Di Ps- ICD CD- a, M 10.C, C

to ~ ~ ~ ~~~~~~oCD ~~~~~~~ -. ~~~~~~~~ (2 ~ ~ ~ ~~C2Cr1 ~ 'Di 0 0.0a r*0 I I=ct a 0 co*d*-t (2 CD 1 p - Di D - -U -0 0M C

0 0 ~~~~~CD- riCD~~~~~~ s- C.0U Di s-i CD 0 r c3 D c C'0 S--C 0 I CDr~~c~~rC CD ~ CD , C CD 5- C ci .-' '. 3 ~

CD CD U CD ~~~~~~~~~~~~~~~~~~~ s-i ~~~~~~ C O ~ ~ ~ ~ C r'~ 5( Ur ri, 0 is -_ c 5-- _3 s-s Di Su Dia CD D CD CD s- CDa

ct- ~ DiD I Di - - CD s -.-0 a a C D

Di 5 . u P ~ CD s -. D C 500t 00 0(2( CD a .t c-· a IC ,a ~ ~ D a * U CD C CD I O a C cr, ~ ~ C CD v5C CD CD s-v (a FP

CDI a-- ~ C CD I

I s I

Mi --.

CD Oa~~0 CII~~~ 5- (2 CD- Di 21r r P a I- -C

-VI O.~~~- Di .~ M

* ta ri0 1 a ct CD D

c C CD s6 5 -,,a ~ ~~~ ~ ~~~ CD L-5P DPI ,cc a C ~_b CD s- -.~P t a c rrc

a ( o a n, is- D 0 '0cIP I a ~~Di CD sI 0-a kr

a ~ ~ ~ c rs- Di D iO

CD ~ ~ CD *·I c tIa C

II~ ~ I ,a r l m a i c a~~C ___________________ _________________________

C)

00

M

a

Ms

0

6-4to

cw

M o

-- e(dm0

cL0to c

t-4

0Du

to

"-I

C0Po c

Page 130: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

I iCDC =CD C. o0 De M /

coQ. VIDP MD~ - ~ ~ I.

-. "' C 0 -. s"'¶ 0 I-.'- .0 " " ~ m,, - '~ D f : ~ '"I II -I a Br· a a a I ~ ~ r- r* a r· I r a a P, cM E a,O 1" .. CI. C' 0 -:- I B 0 c. c:

( -o a - I'M- :C iCD Oa M C

eV Q eV ?D CIM I~ cc ~ ~ ~ ~ 0C I CD 0 r~3 0 3CD cI. cI I-c Cn 0 LOi

0 - M M - D I 1 C

c. - 'M M -- a

M A In 0 P

1~~~~~~~~~~~~~~--- el I= M,o a C C· == 0 0 C I" " " -

. . ~. U ~. C-. 0 D- c-- .II co 0c a 0-I U - .

a, o ID ccCb 1 C M CD C- Cr cc a -z M C, M a e- M "

CD't~ 0 ~ J DJ c-I-~I CD C>V 'ICDnr · I~ ~e m I a De C- 'A CII 'v M

~= v0 c- ' " ) r. c . ,I - , c-I" CZ ' l C D - C IICe . -</ cQ 0 t E el W C> C 0 C b cI j - I0 · 0 .- CD 00C11 cra ,-r B U CDv-I

~~~~~~~~~~~~~~C-ca 0 ei B *-w ~~~~~~~~~~~-w- 0 0 0 0~~0 =1.O C o p ( -~' lu) r· ,· r· II :I3 4 0c P, Ct cI co co CB C. to CD Ca I

0 ~DB r D.. t M - 1, 1I-Ieo c.I-I I D D --I . a 4 C.I0 0 C-ISW-C

¢'I' I~ 0 I CD I-. CD °'a C o a a on ( w· 013

S I I c .0 - I=Jco C- w

C 0. r·- 0 aC 0 .

'*0 '0 CaM C->t 0 -c--'i 0 CIa~~C B a C 0i _cI-LCD 0 C c-. _ -co C-5 M t= 1~~~0 0- 0 D C - DL

[L ( I , C I=G, cr fr, 1 c1 0-CM I-'I co c C0 -. Di Di CD Di- 0 - c-I D I Ca

CD Di C 0 c -I 0 0 CD C a-I a n r C'0 C D 0-C -CD 0- 0-C CD· I-4

'0 D CD CD 0 -C > I CD c-I-Di Ca P( 0 CD B' C t 0 C CI2r

e. u 0C CDCDi CD 0- 0 I-

i Di C 0c Ca 0 Di Ca 0Ca rM OM CIf , ~~~~cI. I-c cr 0- Ca Bl Ca CD a. - -. 0r C' CD 0-. cD '- o c 0 C ct-= 0 CD I-

0 cc M w CD B 0 h C 0 I I- CDM * Cn C-I M. D I-i/j a I ru a Ct a ~r a P, 0 0 C c--C I-IC 1Lt

.0 0 n a C 0 I coa o z-

C=0a , -C- C-I- 0-- 0 0 11 1c )

0 I a-o3 'c~l Dia~ocC 0'00 I -I~~I- ~ 0 CD 0 l-, - c-I Di 0 Di h- -C' C c -~ ~ CD .- BC .cL *c 0C o0 01TP c ·

0 C 0o c CD0CAaII-¶ Dic a- CD a- I --~r 3 re Ca~r U -I- '-I c-I Cr-b4CI h I (20(~~ ~CC-I CD I -- -I c-I Di CD c-I 1- CD Pt I P -(i Ca CD Bn 0 C r I a Di Ca 0 cI I 12 -t a CD Di 0 0· 0 *- I-i 0z r ·

0D - I c-I 0n (D sOC CD tl -I- oL Di I I CD -I s C

n LO CI l, Di P c-I. CD , 0 t 0 b Di CDLO a - CD CD bE D"' 0~. 00 cCa

0c sO CD c-I CD CD '0 c-I '- II aI~ ~ CD I-- I-I I-. 0r Di 0 ', II '- s'0 I-' Di I-, I-. 0 ~ c---a CD P II O I

B U r 0 c '0 1 c-I '-4 0t Di 0 D'I CD Dir - '0 0P -4 I , c-I C-I. '0

Ii ~ ~ ~ 0 I-" I- CDrr 0D CD '0C9 a-'- c--0 00 er

c-I (2 0, ~ I r c I '.0 I- r ,ro i *~,

_D a-. CD '10 h-. c a 0 DI

*s CD 0- c Di 0 CD D iIIC _____ i r

a Ca Ca ICD M

N p C · 0 C-r a ·-- P a' 0- I -h CW'0 C,2 1 ~ - DiC CI-PI CD 00 U I Cb Di - Di3 C C 'c '1a ~ ~ ~ ~ f- Di B0 CD Ca II ' 0Ca'0 - 0- '0, Di c-I II (tI-CD D I B c- I- rI- I- rO c-I C12 CD, 0c CD o II a cCDY ID 0- Ca~ Dic c r I Q

* ~ ~ 1. f-ICDt 0 bP

to a I ~ ~ Ca 0 C - n c-I. cCD0 I C Cacr 0 I. CD a

Di C12 Ca '0 '0 '.0 c-I -tIa ro a=P Ca - 0, .--- * 0 r

r ~ ~ ~ ~ ' CD CD a 0- · I tC ca (' f-I B -*I

r·~ ~~# I- 0- 0r _D CD1f-I- 1 -~ c-I I IC r

II 0 _ _ _ _ __ _ _ _ _ j

Page 131: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

I

a 2 I, Z -, o Ca o F - a

(2(2 b4 (20 ~~~~~~~~~I I .- - I:: b I:I ( 1~a' O C)(2 F9

0 '-c a' Ca 1 %t a a 0, 3 IF-I-r c 0 a F-I Cla C3IC a ct 0 rI, iI-

..-,I~a~ C -C0 C "CO ~~~~~ ~ F-I F-9

'- C iIA o a a (21.a

iC~~~~~~l-0 1 a. .'~ (2J ', Ca I' I-* (2 a, cn Ca ca ., 'I, c -0 1.00 I-i 0 Ca 0-a' 'C to i 2 o a.C r cD=C," Ca '9 FI c* 0 C 0. a a' I

l-Caa c e 0 . c. a '- a 3 . -0 : ) CM ' 0

o 0-Ca 0 h0- -~~~.a'0~M 0 ' 0 r C3 C * 0-C O - . ( rI3C

n--~ 4c, - I-. c aC1 I- Co M1 ' I C =

0 F = . - M = Ca , ' , M-. to CM- . V I- * to cm. a' Ca 0.= FI *-ca '~ -.- 0 C F-0o Q ' Coa a a'.a 3 Cc r· I - ~r ~ -'-· a .. w 0 a'C", Cm.. C C C a I - -0 0Co c t .r C a Po a a aII ar CD CD-. u. .w C:,

* o0 o a' 00 F-IO- a F-I c cj 0 J-C C*C

':-, 0 U ,'1I--d01.Im !W - C Di-Co Ma o c- a C 0, , 0= cr Co M.- C In I- F- 0 M V

~~~ ~ ~~~~-. C,^ "I, =I C-, CDtlto ZZ C

~ ~ -. 0.. 0-, i,-d

~ .-. -I0,'C a r ~ -I I-,,. 0: - ' *a a C., , .0 C C Ca ' iC D I- c C- U '0 0-Ca - I- 0 bI 0 ' C)- am wC M ' 0

a' " a a. rQ IF - 4 U a,, , Ca- 0 I. 1 " C C I, V , ,

'-m ,- , U 0 - -C-I0 QJ 0 0 0 '1 a' F-- a.. 0 Ca -C'-=I 'V W I U 00F M -o M

C -o- c a- ', . a ,, I ,. -,-

Ca -I C *-y . - C:31 0 a.. a C~ ei ,I a 0- '.00 - 0 a. a'a a ct a a ro oQ cm. -a Co - MI W Q A

U - C," Co >- :

a' C) F-I- 0- 0-~~~~~~~D -.

I-- - C -U dl9, F-I I:: 0- 0 ---

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.1, 0 ,,-,0 I-,1 -,m I c~~ p, I C , ~ c.

II C 2 - '--,,- --I -- 0 . 6-. 0 0 CMI - CDo O a. a' e~ c.' a C - a 0 t-' a-.'· ~ p, - (2I 0 0 -N4C 0.C a N a' '-i - · 0' 0 (2 a C) a-- a' I- = '-- -- , = 0 F %0 -I --~ a 4 w a. o CD 00 -. -..a C)

.a' .~ Ca- 0- 0 .-- 0 ' a C 0 0

D. a,"1--Ca M- - C a 0' I 0 - Oa0 0P -a~ '0 '-c -5 0· V- Ca 00 0'0D

0 t q1 0 t 0 · 0- 0 0 I- 0-'0 -q cq m- 'a-F-I '0 c II a. F-I 0 -f 3 S 0 QJ mO r

- = a-- : 0 - a- C '0 -Q-C .g 0.-. I-

0 5-" '-6 110 1' 0 :1! 0V a' a' a-... a.. .M. 0Ii Ca 0+ F? - Ca 0 0 F-I Ca 0- 3-Ia.I -- 0 0 C

Ca 00F= t- ma., . 0 0 - ' (a 1- 1 0 C,~~0 V-(2 a- c-I- Ca C)~~~~~~~~~~~~C M '10 '-V M -* a. a

0 C7I 0 0 0- 0~~~C M to C2 0 W. -Yl CDa CA- *·- Ca Ca V.J 0. '-, '-5 0-1 (2 If I- 2 0 00 0 -atam OCa 0 (2- trt 0 -I 0Ic

0-'I- Ca- 0 0 0- ( ' -- =. Ca aM 'a Ca tow 0 QJ F-It '.0 Ca 0 C--

0 + - 0 r 0 -Ca Ca 0- a '. 0- '4 0~ (2I-V a3 a-h I C a a- 0 - 0 C a -

' -00- (20 F-eC I 0" to c = r Coa C2 ~ p C t1 Ir·a Ia C

ccCoe-M- m- ( in 0 a a c- 0 ta I , C Co n n C6 0.C-I, - I- n ( F-I- a M 0 0 5-- c- I- C 0- Co cCcc3 c C a 3 n ~ aa w Co Cacta

C . M F 0 '-0 0 - 0 - a m aCD C ,. - ' C - r IV w- 0 - '.0.

" a p= C- " - U' C12 a mV a-" a- Co C C., aJ (2 0 e t CD 0 0 (2 ( .

C o 0 m- 4 0 0 " CD- C - -MO (2 C F-i '- 0 o Ef c a Oa a I o r· ~cerg r w a Ca c -C. I- C Ca * I * o 0 aCo I M ac M (20.'V0.C a.. (2 ~0 0000 Y C aO s C·

= I-hO cI=-Ca

-. F-'- 0 0 0 m~~~~~~~~~~~~ ~~~~- -1 -C - .a0 0- Cj F-I- Ca w Mc Mr M M - 0 0

C'2 ' ( FI C cMM-=-d - 0 a - Ca 0 F-I-Ca 0 CZ 0

C.I- c a-0 C 0 (2 cC CD Cx. C -0a~~ ~ 0- F -~I- 0. 0- 0- a-. 0l(O 3 C

'-w c0 ~ 'n - m C i a ' * a- C toa-> Ca -, C, - U 0 0-

C 0 0 C 0 , toC. - Icr P C·C a c t 0c 0- 0 0· 0- c-in- a'c· 1a aP ct ,r 0 a' -~ a' 0 0 n -. 5-' 0 -r a'r

_ 'Vr - ,c Ca (2 rI- 0 o 0· F-I IV a' 0-"c ~~ 0- 0- C 3a a '0 0-1 F-I a. a- -. S

C, M F D 0B C- ,· 1 a I ,~IC- mv IQ I

0C)co

,-I

g'm0 C

C 0lDo

C>-

0 lo

0

'mC Xt o

-C-

CN. 0ccCo

'-.o0

'-4

.4es-- _C)D

Page 132: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

'D J-. c: u.4 -C D,~~~~~= g CD CD 03~~~~~~~~~~~~~~~~~~~C C-'*

CD PrC 0 CD~~~~~~~~~~~~~ oCO ~~~~~~~~~~~~~~~~~~~~~~~~MC, M1-

20 =0 " 0 CD 0 0 t-. 09 t M co 10 ~~~cl 0 0C0

C"~: = {~,; 'o 1= C0 C:A {'D '-

C.e CD n c"-e C., . Ch I M-'W. - *. , W C", W 0 -4 -

= :- CC D 00' = 0 t I 0 3 0"' - -0 3 0""- Igo 1 I0 D 0 30 ID- I -i= CD 0W2C- 0 U2O 0 CO Co B t-t . M CD~C 0 W--a co 0# (26 '~a 03- W C .- 0* CD W3.C = 0 0 > 03 0-( - '1 I'd CD COC . - 0- (~O M I

k,,.,-Y ~ *-4~ CD 0 ID · c.o - : a, l- A.-- =0 -U = p 03 *~~~~ ~~~~~ ~~~~~ ~~~~~~~~~~~~~~~ C'O -0 U - 00~, 0 0 0- -- '0 U~~~~C 'M W co O C 03 D 03

I- __Wet co W C, MSu C IC C- co C -0 CA n M 03 O - O DC - so co (-~ W' 0 CM. 09 '% n W,-- , 0 " · ::0-3C 2(4 0 -, 3~9 - 0~ 0 co I eVi ci M~ z 0- C .0 0 100 D- C t 0 0 'v - CD C> C Ca D a

(2 us 03 0- 0 =1 0 (2 0- - Z - 0 C0.D0 C~~~~~~~~~CD~~~ ~ ,'~-1 0 (~ C 00-a , ' 003030 42 0--CD. C 2 -- C - t'- C:,

(-9.0 -- 0D * -- 0 2 C0 n 3. 0 CO -00 -c 0 0 - -~ DC 0 ~ CD (D

me :. D.- CDC CDC to0

93 co -9 3 0- CO

CDC ' c- EN' 0 e-¶ 03 '0

C' D C-4- -. D 0 C - C-2 O 0 ~ Ct

-C me CO 020 0. D 0(2..-C-9.etr'C.,C00 M C- 000 to

- 0 0 ~~~~~~~~~~~~ CO U~~~~~~a- I CD M-- ( --J CD~C W-W 0

0-C~~~~~~~~0--4 '0~~~~~~~~~E C 'W 0 9C, ' 0 C O. -

CD 0- (-9. 0 ~~~~~~~~~~~~~~~~~~~~~~~~C- 0 0* ( 02 WO ( 'C

0 C W C. cn 0 CCb co to7 e-3-- C> ' -C O '- 0

as e-0 C. -1 -

* I ~~~ ~~~ ~~~~~~~~~~~C M- CD 2-0 C

I *~~~~~~~~~o c L -C 0 3 v-

0 CD 5 0 0~~~~~~~~~~~~~~~. D O C> 00 C 'W C(20 M C,.0 COD toM3U M- CD30C o

;- I2 D 0=- C-to CA C-

030 C~~~ I '~~ CD-CD ~~~ 0-COCO(2c4F'- c2 - o C,,- .

c- WoC ~ CO ME'

C~t- Ca M0 Q- M 0 CD oOM0 CD CD 91~~~~~~~~~~~~~~~~~~~2C. t C.03 (-9

_ U ~~~ ~~~~~~~~~~C CD -0C

'9 0-~~~~~~03 0~~t~ c o C02 '-4 CD CD '-9' 0~~~~~~~~~~~~CD CO-a CO 2 0

(2 (-9. CO 0~~~~~~~~~~~~= 0 - co D - - I

'-I a0 COoU039

;C 03, ZR c CO ( c ' co a -9 C., I

C- to t a-9

_ _ _ _ _ _ _ _ _ _ _~~~~~~~~~~~~~~~O ' MW

to3L

as M00rMlw'

- I-u

. C0

toO-a

L~

02('-4

Page 133: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

I111I

(a (a

Ca_n O

co ;Kl

c, -C .o Ci

oCo M C D oo t o _w' _ 0 o

I ~ (a r (- -- C--~

to cmX o o '-* 0 -- 0 0*o (a e- .' _a 0 a-Ca =C,,ca Ca f-2 . C, 4

t o 0- CD o _

C2 co r 0 00 *

W oc0 CUI~ U sl i r.o _a eb~ o ( c g

C> _ CO C-o a . C-.o eb --

PK CD :S o Ca fp. f-. r C-p.1

> * a z-~( a 'fp =r C. 0

(a p, 0 0. 0-

CD v c am__eco CDX" * .. 0 C- C2 a' D0

_.co CD

C, _ C,(a - "g 0 '.

CID C l' - -

=O C C

0 I Iz eo 0ca2 C ( -U p._ .- O C5a' p 0 f- p.1CU .4 _~ rT D* (a2 Ca (ato o

Ca Ca .0 0 C - 0 a8 _a't Ca - (a O a

(a Ca p.b4

. ~ ~ ~ ~ ~ 0~

ri- w .

. (a(2 P

_

. .~~~~~~~~~~~~

I e~~~~~a 0 0 _' (2 0

D' '0 '

I t~ ~~~~~- aV .

1-

p... a'

C- (a f- Da

l~~~~~~ 0 (a 0_

I~~~~~~~~~~ * 2 lI j_ r~~~~~

l ~ ~ ~ ~ ~ ~ ( O- -I ~ ~ ~ ~ ~ ~ ~ ~ ' -

m w r =L.la' 9 I '0 -0 a' -

co0 C a f-p aO (2

( O~

0

C" _= * 4e" C> *2w co C t> toS ·a , 0" 0 a'" "-,

C> _ w (2 .- o-o o C (

~--h W. ( D , ,,-, . ,s= n C-= C C

to 0CD to (' 'C . C

I-.n- w' I~. C, w 0 IC w0 " C , oo w :C,~~~ C : w n n b C2D I- ' 'o - ~ ~ CaX. (

co~~~~~~~~ ~ ~~~~~~~ I. Mw -,-=

(a 0 = C> -. ( , -I ' w

C cO 4 a (a 0 Ca _o f-S C o 0l c_ 0v O~ O a' O

C O C -<D o.' c>0 _ C 1 RC

~~~. m

~-. ciC, C2 co C-- =- 0 C f

- O_ D co Z n _ C,D -t =1 D C- D - C I so 0- 0. 0 ( 0ft, CD a, le n0 CD a' 2 a, c (-

tv~~~~~~~~~~rtI-P - f- 0 c cv oC2 Pr _ - -* a

. rb co _ w o CDnV V- r tl CD to C13 0C V co '__

C , _ , - - Oa,- O O OV-00 I (a ' V

0 D f-p. (a - 0 (a Cp.. (a 0 Ca a' O IC Ca-Ca Ca 0 0- Ca 0 C_.( a' (2 Ca (2 P-- 0( P OCa h-a(a f. '. Cp. f-p f-p 'X n (Xl (a 0* 0 p.- 0u 0-0 p.< O2 v - (2 0 0 a

D-0 Ca -(a CaD O~1 Oa' _ 2 4- f O .* p.. 0 (a 0

W Cn 2 O | § CL fD n~~~~~rv 'O

ca~~~~~~~~c

l~~~~~~~~~~ C> CT tf0-~~~ ~ ~ a'l °l F . a f' ' U 0 0.a

. C _4'.0 h-fP ~Ca U Ca ''~'~(a ~ ~ ~ ~ ~ ~ c O.C a 0 a ( .fPa

to0~ ~~~ ~ ~ ~ 0- Cafp 0X1f'* aC

!::-0 0- f-D

a (2. ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ .. (2 ..

- ' :

0 f-

.a

0

O o4

p.. .I C. SC",.

_ __

Page 134: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

--

4~~~ ~ ~ rl Q, 0-. 0~-

t- . -m I-

if t.411

Cm .. . .1- M I:-C. II o.

C~~~~~~~~~~~,'T s , O ..i (--

4-s.. ~.. s-i-f .¢..4.C~ ~ ~ ~ ~ ~ C~J 0 Cfl W -0

I ' II I II co 0 '- c

II 0- ~~~~~~~ ~~- -

Urb 2 M b .- -0

-a 'at- C-,-a-tt.0' 00 '00' 0~~~~~~~~~= ~ O0 C.t

0 CD D aW 0 C> - -- I~'I- 0.0 0-00C, ~.,iim~~

to . -~ ~~~~~~~~~~~~~~ c o n -0

a ~~~~~~~- ~~~~~~~ 0~ w I C-~

00 CDsi-D C

a- s-i- --C, a - O -M."V = , 0' 0 0 sAl a n - to

04C C, '0 s--si- to - 0 0 0

L- n to C-to t to M M M

0- CD~~~~~~~w CO I, -d C> I rz ~ ~0 ' coi-..C m 0 am -* D

CP 'OJ - 0 -- i- #5 a C.-.to - "- o co -b -(~a-~a CD C M C0 - 0 , MCD c 0 M i - s-i--I

-4

-s ~-. a o~ C, - to - s-b o 50 ' 4too -- "> 0 r 4-1 0 - -. 4-5 -. 0- 0- to C ' M- 0 t CD o C.. 0 0' .-. ., I -5

0 0.40 CD C.0 -b 0 t cosb 0 o 4- -- sito -=.a mm '10~ ~ ~~ s- a CD IV - u o C

M M~ la b 4r-a ~ s- -M' 0 si CD M to la 0'e 5 4-b ~~~ 0-~~co C to 4 - ~s o ~ I~-i ~ i-'~--. tto=- n . -s 0 *

a C~~sO ~~ s-~~ i--to '- ~~~ s-sm in to~c

Page 135: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

III

IC~~~~~~~~~~~4

0) 0 0 C-

M W Cc C", CO n v- Es Co C-0 ~ ~ ~ ~ ~ ~ ~ ~ ~~~~~~~C C.I gv--M. 0 ='., Ca ('3 )

gm~~~. =: CD - 10 n CD' M nCK- CD _.- C - -C- CD *-

~ ~-~~~~~ ~~~~~- -- W n comCD

.a --dC, to CD 0 E r s- _j = 0. " -' 0. 0-CDa oo r2 lat 0. co CD C0 1 C90) C*. n M*lU

'~ D C ) ) 0 0CD C- V9 ;-CD 00' )00ooy 0-0 C", Ca -0 W~~C 0 - C 9n. C CD et, to- 0.(

0 co ~-W ~ CD Ma - 0- I-. Ca .- . ) 0 CD CD1tCD 4 C> = 2. CZ 0 ' 0 Mc -CD-i '-s C"0 '1 W Xco -

-o (*00 C.~ (*0) CCD0-to' I. C, r mC

4--'- ~ - ~ ~ 1)20)0) 0 -u C C- 0 CD

0v--t -~ 0 0 0 CD 4-- ~. C D--' 0 Fl- c0--

=I- 0 0 CW - 0 o C.D PC DS= C r 0 C. ",C-'a C D 0e P

tW ,W) C -5 (1 CD *0 ,.D 0 Ca C, 0) -. 0-C CD '. _ C*l~- -- cow M =. 0-4- 0 CD l IC I - =. ., Ca .o- =, - 0Ca -D Ca 0 S - * m (* C '1. 0-'WI

to I .-h~- 0~~ (0CD C

0)-- ~~~~ Fl0 0 ~~~~ C~~~0 M- M w =0C O'. ' 0)0-) . co W*~~~~~~~~~~~~~~~~~~~~~~~~~~Y C ... C',,r ' 0 0 0- ~~~~~~~~~'. 0 CD 0 4-4 4-" 0) * 0~~~~~~~~~~~~~~~~~~~~~C in (CD = - 0 CD 0

0- 0 ~~1 CD 0 (*-w 0) in 0)-~C co 4 MC= MT M W. C- ~ 0 C

00- 'e' 0 , , 4 0I in ' -- e 0 -: E. D 0 C' > 0m. .- 00 ( J (D 0' Ml 0- -% -4. in 0

k.~- 0- 0 4 * In 0) 0 c0 c Ca ' 0 CD - I 2

:("-- ¢'3 - CD I F- I - -- , 0 C - C. .CarD s-i Ca 0 (-2 8 .-. ~~ o- , Ca sCa tct, 1~~~~~~~~~-. WCD o t

-C (D (D ~ CD (D C C O Ca C 2 (l-

,=

0-C 04 CD C-* *~~~~~~~~~~~~~~~~~~~ C 6 - !D '0 CD4l- 0-CD (*1 Cl~~~~~~ ~~ ~~~~~ ~~~~- IC CD

C IDCa4.0 0- - l- - 0: CD

0) C2 (* ~~co - 45C Cl- IC CD' 0) C II

tor C > - -4'C -CA CD' C.0 W M W ''0 CD 0 0) ~~~~~~~~ 0)U 0 i-I- - r-9- Ca 4-9'~~~~~~~~~~. s- CD * -

L)0 9i a 4. D 0 i Ca ) CD - CD 0 Ca II 11

co C-'I 0- C, W co0 C 00)i CD 0-I Ioo C>~ 0 ) - Ca C1 '0 0)0 0 0 i a

0D C C i *. i 0 49 W) 0) eD' CaM C12C- C-4- CD ) D -' ( 0- CDa s-SC

C> W4 C0D 0 C -- D CD C-- 0 0" - CD D. 0)0 0 0 D co i -i C

6. CD 00 0 -3 0 M0cD(l- a,) -

(.D- n l-0W-" -0 0- 0- F- CD C> Z $ 0-; C - ( D 04vC- 0 F-- -1 0)0 Ca C CD) ( 0CIO'

- rl- -* Ca 0 CD 0- Cl- 18- W~~~~~~~~~~~~1 CD 3 C

CD ~~~CD C2W) i - (*0t3 i!0-I -0 '0W-FC 0D laD0F44W-~(C

0 C> 0 D C W4- CI

S. c me C. - Fl- 41 C-0 r- to 0 109 C Fl -

0% Ca 0 Cr D Cgo4- -C. '1 C, IFl-C M ., I'lW C. 0 0 -. 9 4 CCD C- 1 3

co~ C, sC CD Fl-0)I- M a 1

00) CD .- 0 CD Cl- 9--~~~~~~~~~~~~~~~~~~C- o 0 0 . ~s-W (0 C=M WC 5 CDZFco W * .. C M #- 0

v -W fr CD 112 ~r 0 C -- 0) - .0C _ 1

C, CD in 0- 0- - o Cw 0- (*4-" E 1

CD .-0 0)~~~~~" =9-4T " . 4- W'0 s-i '-4 0- CD ~~co I co5 . C- 50-I

CD a-c - to 0 .0 a '00" i b0 C , 0 0) 2 'D - C, 0 - = 4- -

C 4 F- i CD 0 n Ca CD = 0 -I-coM

CD Ca 0 Ca W CD~~~~0 I 0 M 0 CD M 10* 0W Ca 0 Ca~~~~~Ms " 0) = 019C1

0 rl C3 0-00)0 C,,"'il* o, 0-- C

C0 M' 492 o D F-0-s1 4".0 0,i CD M ~ 0

0- t' C - ( 0 D0 e-4. Ca 0- -)20 . C) 0 C0 D

- CD Ca ~ CD 0 '-~ C-,- a- to - s ,,,

I D 0-00M CaC 0 0- --a M I= ~

Page 136: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

c.o Xunr

CDD O

,-. 0 tI . , CZ CD~~~~~~~~_ _ -=_- , ICD i,.~ ( -I- .* - ' " C1 0_--,-E.4 Im t,. " ~ ¢ °co

n~ B'A ¢ " ,.1 0 £~ (U

Cn~~~~~~. k., _ C-f..mmA c-I E 0 ( c cO 0co - . C D ( Cc_ ~ 0 1 :: tU 0 CD

W* Uc-iO -b

0 (I -C O~ ", C CD 0

Ca..~-. .( 0. 0.. 0 W (-4 _-4 a W nr to (2

c~~~~~~~~~~~~ ,-,1 ~ ,.c~o(ocI-a_ a tv 4 C.0* a (2 C CDDto to _ =C~ _ -. eg

i _ C-l t: I ' D

c-I CD c I- 0 .BA 0 (2i 0 Cf O 0CD CD-

92~~~~~~~~~~~~~~~~~~~~~~.c- b* BA CD I

coI X = b e

O . tt CD g- BA ( ' -. BA,,

D 0-- b * _tCD CI2

(M. Ca ( -c-i-.u CD O 0 CD

DU b -w X b Ws- - . 0- E 00:sO

~~ (U BA CD B~~~~~~~~~a -- O O "O fD ~~~~~~~~~~~~~o =c -' CD C

Dv O O4 eelD",>

CDa C-t t7 (t(U C c-i cr I-

~~~ 0 ~~~~~~ ~~~~ CD BA~~~~~~~Cc-I (U ( CO BABA ,_I- c- D( toI 0 t_ S 0D 0 CD I-

n cm. O 1* MoC i- (- U crI

" Ul - _= Vo O nCX O - c D1 -I C D c-OCD - . .n D OO C _ -0 1 O 0 <* - WA B m S -

C M. =- -. P0 W M = ( - 0 e

ta D ~~~~~~~~~~~~~~~~~~~~C I a p0 - *A tr(< so *.......c-i-. EU (2 td _BAO(2 0-(2 C 1 0-c-U c O - ( D B OC O g 0 cr ( -_ cC - c 0 O X 0- *~- c-X 0-C D- BA (U c. BA Ds _ _> @ ~~~~~~~~~~~~~~~~~I- O A t- DI -_ _ _ ~ cn C 0.- C t:- - 0- 0 _ It

0 (U - O ' c-iO C D D CA co C D -co t= (2 t-0 C, 0t, -* I IIr C= 0 o (U 0 C - ( gu C> '

(U CV - .- I-- D- D _ 2 (U cc 0VO '-- _ Ct - C-B (U BA U IC ° n - MtC1 WntC LC

0 i c2.D -Z : 0 BA O 1- fD (U- &- CD c-§ °- - 0-.C, cm c. o ( 0 ( B I

~~~~~~~~~ i ~ . I 0r > 0. U c-i 0- (U (IC I- ')( . ' to el CIC OD _.CC CD C. C

I _ o Av Ap ' o(U II C>- ~-- (iA U CD (o _ BA S0) '-0 C '.

(-2 _W nC-0 cM C- X M

t= CD tt Mc < B c-D

C CD BACD 'BA C o co (U C

CD D

0- CD 0 W 11CDL~~~~~~~~~~~~~~~~~~~~~~~~( c-u c--BA

CA

o

-2CPOMX _

CO

C

..

~moon. I..al C ll C

Page 137: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

III

a 0 'a0 Go C9 -,(2 a-1 (C a, ( a -

Ca (U Is -c

me- a a -.

~~~~~~~~~~~~~~~~~~~~~~~~C - C,'-

.~ 0~~~~~~H I 3 CC -'1 aCa a, M 0 (-2 14 '100303C. 0= -C 0'o> CD 0. e2 C ' 01- . , Ca w-i 0 0 Ca 0*- 0 0 ( 0 a MC a-- 0-aa- - 0 0 "- I to MU 0

C>. M' CD CA = D C.' M c

r- M C- C. C- = w= ~" co -o 0 O C - 0 a O 0 0 - C Ca k.1 r 0a--r,. -. C- =-'f- 0 .0 C a I

'0 ~' 4 0:: 01 a-: * 0,- 0-'0, 0- 0 U a ae0 '0( 0-OCa f (U -c (U o·aa (U (U 03a

M. r 0 0 D 0 -.C. a -. a a- (2D I '0Ca Ca D0 0 ( - '. (U aa ( a C

o~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ =~ m m, 'm mt -- M 1 .-? C= co,,-- r = -- -C C C2 0 tb C- Ca a M w0 0 MU 00a 0 C-( DIac(U C aa-. 3- U W oa

~..m0

~~~~~~~~~~~~~C 0 co $.. :~ ~ e~ ~ ~~~~-~(' 9 'o (U 'ora 0 -0 a a- '--i i Uc g

a,-- 0" ,-.. -a. ': -. I (2 0 03 0 3 a..- C, ¢0-'1 (U 0 1-4 0 1-ft 0-0· 0·1 cC as a-a-u I Cb aPc a I a a a~~~~~~~~~~~~~ C> w w a, C- O I ar- 3 a LI Ccr ) a~~~~~~~~~~~~ C C, C, C> n. ,CO (U a a-' a '0 (U C a- a 0 r Ca , a-' a-'- a-a 0 cT(20- a (U '10 (U r10 0' (U CV2a11

CO CD( ( '0 ( 0 U '- ' 0aa- I030 0' 3 0-( 03 a - 0- 030 aar - I .40ca a-t Ca 03030 0 caa 3 a-'-ta '0 I -a--Cv2L. r10 co0w (2 a a- aa - I. co CD CD 0 a 0 ( (2 a '( I -(U '0 a-Y ( 0 0- (U (U p, a (2 '-, 0 0 I 03 a' C~a a a (U (U 03 a -, a-a - t a-I (U '- r-

vra I Ct a a a c :~i-. -C00-TL * (2Eua a (Ua r·C a--ft co a-- =- ac -yC,

a-0 C0 c un t - C E Ca c-I a t1O t E, Ca~ a 0- a-I

( C co I I- CD 0- 0-'1 20 0 Ca aa- (U CD

-(2 M* a-a ( 03 Ca-~ U =- 0 C (U-(U(2 a ' m 0 ~ - a-' i-ct r Co I a cs a Po(U CD Ca 0 = 0

C - (.4 . M 0 M co

(2C- C C c ~ - a-3 -5 a Coa M, - t(U (U 13 0' 0 a-. I IC

'0 0 031~~~~~~~~~~~~~~~~~~~~EC S 0o a 0 (U C'~1 (2 a-a-.. a-I (2 (U 03 (U -, e0a a-a 0- r a-, .. a-I Ca .4

o 0 (U Ca a-. 0 (U a ( CD0a-- 0 Ca fi 0t '10 f1a

( (U.4 C ( - 0 (U

CD C=F a co C- o0- 0 a-a- 0 c ' I a-a- 0- (U(U.C (2 (U 0 o '0

- rI- 0- (U rt - 0-- 0 '0 0 03 C-Ca

CD CaC *0 0Ca c

C- co~~~~~~~~~~r · c

cO BI ct a P, C~~cs C a a a a a I a a ·- ~~~~~~~~~~ a a' 0

'.0 030I- (2 _c 03 a- 0- - I , u cc - (U a-i a-. c a-I ria-0 3

a-a- (U (U (2' ar a- Ca a-a D a'0rI" 0 a-I- 3 - a (U0-0- a-i -0 ~ ~ ~ ~ C a-a 0 (U Ca

(U M-0 C ( n C 00 M(U . t 03 '-1 (U 0- a ccr I .4 (U a-a (Ua 00 0- (U a- ctP II (2

a-. '-0- 0 Ca Ca C C t o C a-'- 03 a'00 ~ a 0 ( a-i 0- I

a-¶a 0c a-. 0V 0, (U . 0- '0 w

M (U (U 0 a 03 - '- I t C cC(U g ., 0 1 aZ I- ( rI Ca ,.

( C C I" M coS. C-, a 0 a-C a -- tP a a CO a a ) or c: c~~ -- 0- 0 I

c 0 I 0 aCI -CC Ca (2

CC 'A -,a 0-a (U a-c 0a, a a-

IDa- co- Ca - -2 (2 a , a a a 0 ( 0

cl M l C c , = Ca-I-a- 0. riO aa a Ca (U Ca (U 0(2 3 0

- a-, (U a-I i Ca a-- 0- 0 a-a 0 0i 0 0 I a C'-a at CO C- a-03 0 ai-0-(2 (2 0(Ia- ' Ca a( '1 - .M CD M0M (2'.Ca 003 a-ri -0 C D0 a 0- (U (- a ( 0 a-I-Ct C a. a-I- C 0 0- - a .a-a (U~ (U 03 (U a-i -( 0 0 t a-a (2 (U a-aca- 0c c a-I c- (U a-~ a- a-- (U 0 ri- Ca 03 . ~ 0 ( 0-a fl 5

0-a t Ca ~ a (U Wa-a 0-0 a- ai Ca 0C W0 003 1Pco Ca00 .4 0 (U 0-( 0-- a-1 (UI a-I 03 a- a - 03' C a I-l (-a '0- '1 cc Ca Ca 0003 '-1 0 a-- 0-( (U 0- 0 03 3 cn .4C ~ ,(U - Ca 0a- aa 0 a- 0O 0 030 a 'c0 '0 - 0- (U ( -'-'0 a-(2~ '0a- a-0- 0- (2 a -a- ai -( (2. -m03 o ~ P a- a- 0- 0- 0-(U (

03 a-'-rC (U cc 0 a-~ ct 03 - U a 1-4 03 a- 0 0'. a ai-'0

-0 Ca 030 -0 1''-1 a-a-03 -. (U (U (U ri-0 0 (U (U (U Cca0o , 0. 0-a 0 - a- - (0 - '10-' a-' 0d .4 0- 03 0a' 0 0-' 03 ~ a-i ( a-a a 0 '1 03 0 1 0- a- -( a-a * a - -- a-a a a-a a 03 a-- '.0 a- .0 (Ur 03. r a-' 03 0 0- t- 0~ a- , 0- a-- ' 0i a- Ca 0 a-'-'0(2 a0 (, 03 (U '-1 ' a o Ca I-ft 0 U 0 2 a- U C

(2 r0 C a 0- C a Cae 0- (U (2 (U~ C at 0 c

0 0 ~ c a (U 0 0II CD a a C a * (U 03 a . 0 -t a rr c__ _ _ _ _ _ _ _ a r C ,I tc caan C1

Page 138: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

V 0 0to

to ~ ~ ~ ~ ~~=~~~~~~~~~~~~~~~~~~C

n 2Dz C CD a W M. 0- 0, C-- (2 cOCD-,-ftv, b = :I =

IC ', M- :---. . . cC, CD . -- l" 0

0' 0n gC~ 0~~~~~~~~~~~C· a I ct p,~~~a ' C 0

Uv~~ CVI~ 0 (2 -. C CI.o 0' cVc 0' - C Ir W1 0-CDfM Cco 0 0 C 0 0 to C-*C~0 D CD 07 o 0(20 C , CD

CD ~a r· o a -eC- la W I- w CZ, `C 3 M QcoW Cb tP I~ M M=~( :C = CD -( - 0A MV .CC,," A -:I ga, W -... a W C r·a a a c a a, ( Cd2 rC f

4C C-2 * CO C ,- co 0 0-

(2C . 0~ D C 0 (D W. C,o ~C0 0' ,- I-~ 0' C. ~- a D '0 0 - 00C aD 0- D4Ur 0' 0 -~ C 0 a M 0 0 CD C to'S CD - t4 t CD a

'0 -t~ t r ' ca CD '0 C 0 .- 0c 'a 0 CV 3 . -' . f* CV CO Ct I(DO O 0· Ca aP 0' 0 · d ct Cl CaC q arO (20 CC p U I

CV 0 -. ~ ~co 0 0 - -4 -- C a- - -0 '( o '0 0 0 *ec ri- f.0 0' i C· 0- fV 0- Ca 0 0 0' -. CVb CC ,D CD J 0 C a C C U o I2'0'0

~~~~~~~~~~~~~~~~~~~~c.

ow o 00 -~~~ C~ ~ '-e'- ~ ~ 0-b-·. ,....--0CVI , I e c"= -, el ,," ~W-0 ( -~0 -0 C. M 0' 0 oD 0' M- CD- co 0 Oh-c D -· C a, C~ C '0 CM W CZ n 2 - c SC 0 -

O ~~~~~~~A =~ c" Ia · a --· t - L r'~ $ -. t Ct 3- c

(2 CD -V~~~ CV~~CD d. V ' t'b O. - (2 -w ~1-C9( -4.(- -C a -. c- e l c S e D C= 0 =V CV 0 CO C-4 eb

CV~~ ~ 0 Ca 0 0- CV I CD 0 CD ~' 0 'C ' 0

" m , Cd- C cDm 0 0-' u o r * C2 0' C -. c 0 00' CD 0 C 0' 0 I 0 CC CD C0 (DO rc Ic

tcc CV Ct 1( U - 0 0 a p CD (2( 94 CZ. e-- V 0- (a fc (t CD f*I 0' C 0 c (2 -. 0- CDOc V ( 0h C -~ CD C b- · rt f-f 0 0 I (2 0 C COa 'V OcV o 0 a

' 000-3 0- 'VCD CD 0 U . mCD) 'C'. 0a 0' CV ca Ca I 'V b CO(* -0- r"' 0 b-t C- "COI

CD 00' N _ I C~ CD a iO CV U C- CDC,, co trD PIctC0I

' .-.- , -0 C C

O a CD 8- .4 0 ' 0' C -- 0- bW. c --CD co I= I c1

C , b C- CD " 0 CD'b CD C- 000' r'- 0 to Ca ( t 0 Ca 'r 0-0 C "W 0'

ii i i i ml iiCi 0 C i il · DSI ~ ~ -( O P O ( I 0 CD CD O ,T =I CDa c 0 - am

0'~~ ~ 0 '- 03 0 ~ 0 CV08-'C C I 0 CD -C8' 0' 0c 0'8--

II'0'0 D V b C C C 0 (2 CV 0' (D 0- 0t - . -CD 0' 0'P 0 0

0-Ca ~ ~ 0- CO ) '-U2' w1 0( CD 0 0- 0 Cc CD I c. =

0'00a ar CD 0 CO'O~P aC T CD CD C 0Ca0C 0 -3c a C ;IGCD 0 Ca U CO 0 ' Ca 0 0 Ca W 0 0' Cf Ca 0'a~~ ~ ftC c ~c C 01 * C CD O 0- (2 0 0 0' U1 CD .- -. 0-· 0 C (2 CD'.~0) C-'.00 0 D 00 000' 0 (DO 0 ( C.O CTa a I-cr-Ll c c ~ c I a a I c c·o S . W co IC.. n

0' 0t 0·~ 0' a- -O 'I -t Cf 0 t 0' 0 C 0 0 0- U VC D0 0 00'.0 0' 'VL 0 C ) 0~ 0 CD 8- t1~ 0 CO ?-C) (2 CD 0' CD CD (2 (2. CaCD CD Ur * (20 CD CDU CO * I-(2e) 0- 0-00C 0' a, a -C20 r~-* (2~ (2 -~ CV 0' ~l- 0 MCDZ a - I-- C0 0- C>CD CD 5 M co-C co o 0r C2 ', - 0' 0 CD W CD rt- O O to C L r aCa 0 Ca '.0 W- 0- 0 -e - C C b ( (' 0 0o -SuaCV0-f- C1 '0 (C C',0'1 a1 U D 0 0M W = =O CD M f o 0' Ca re 0C CM Cr t- MM 0-a I

CD C3 (200 C, -1C 3 CD Co 0' 0 (2 cL. CD 0 (D Ca C f a CD, 0'· 0' a-.. rC (2 0' '.0 C-'- - C9 -- -' C 0 (2 CD C. C

I 0 U b--0 O 0-0 ' C 0 0' 0' 0' CD (20' C ' 0' CO ~ coa C..1 coC -a C> CD f I a CO 0 0(2

Ca WC CV I D CD W C( CD 0 CD Caf t Ca 4 C Ca C CD c-1M C 1 co

M- co C. 0- -I 0-. CD ,

Cc ~ ~ ~ ~ - 0 b CCO CD C-W a a' (2' 0' Uc U C-- 0- Pmr .-C 0 C9l bE 0- - 0 8 f 8- - 0 C IC 0 0 -t= ru CDCD CD D CD gCD CC 0 WS. -C>P C· ~L a a O ) If- CT=r P) W CL) C, ,Cc C- , a -

IL o 00 (2 C cV 0-r Cc 0 t C-CCD C. C>:0 00' - ·Ca Ca t-· CD 00 - C CV0- Ca 8-' 0 c r0 ~ t CD c 0 ' C.0 r r - '- DI0 t

(DO C CV Cco -Wv tD -o 0o CD CO,- "0 -- D 2 2 C. e~

' C0 C-COO 0' 0' '.0 '0 C .- cD'VW M c~ ~ r C4 4 a 9wo, A. CD a C-L. C C-CCC, coC-CD 0- - 0 8- 0- Ca 0W CCD co C C C 0 - 00 0.. t U 0 U 0 ,· 0- C o 0- C 0' O C - CV

'-0c C c3c C-CC a'0 8- I CD -- '-C 0' t,-C CD CD (2 0 C-CO C IV W W M COC-P Ca 0 Ca 0 0 CD CD -C 0' C 0-0' C- 0 -h C'-CV -C -c=--- C0- =-0 0 0 CD -= ". - - CD= , r = C2 M M I( c c ·

C, -- e b mc CO a, InC C 0 p C 0c 0 0- V C- 5o 0- ' - Ca - - 8-

0' CD C-C C ~' D C1COU 2C C= 0 D -CD, C- CD t 0 ~CD 2-

0 C-D IC *- c t

CD ~ ~ -~0 CC- ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ '0

C),mMe -

11 ,-

,-4-e.-

co

-4

-¢-C2

P 6'0

4.o

Page 139: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

=~~~~~~~~-~,-,~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~. ·A'iD

(-'I'- C, =r' {~ ~ CD C

..-~- eu"1 , MT w ~., C M.- ¢%( 0 n 1 DiC(U M C13 U C- L.0 i

Di ~~~~~

-, I-1 "0 w 0 cc C w 0 Di J C ,~D w I " -a ¶ D 0

co o ct o C,-o ScoU w M ' b-( DiD 0 CDi 1

(U - (U. 0 (t M1 iD C", W!

CD et - a C- :W '2 (D - 0t - L V

D -C~ eb (U0 M 0- C-' U .

Di (U W 0 Di (Uto Di DiD I Di -

a - ( IO C" CM. c n Cy-. T U Cc D W

(U D as o C n~~ t? Qi =- 0-CD

D -. b. coU i (. 0

C~ ~ ~~. ,.C,~rC. 0., - , CDDi Di (U (U ~~~~~~~~~~~~~~~~~~ Di Di ~~~~~~~~.. Di - 'l- 0-o co 0 (U ( U C> n A Di 0 C

I~) I.*~ ~~ =:"' ~- 0 ~I- M~ : =1 =O"WO IV $ cr r' t' " n cm~ -, -. 2 ~ ('Di Q0 ( M (U ' 0 M(' 0 0('C CCo

=,- ,,,, I ,i--- cA., 0 -c - - C-.'0 v1- 0 " ,- - i--. w - 0 D o C)

-- C = 'Oi D - C ((-

I". (U (U

(U - = 0 '1 - 0 0 Di i 0 (lo( (- = -V " - 0( Dia-0 Di co C. 0Di -( 0 "'O-(UD 'O Mi (C-.1 CP-

0.0 Di (U CiDi( 0-0 Di C- CD (U ~~~~~~~~~~~ DJ~~~~~~~~~~~i (2~~~~~~~~

Di Di C$ * a C- Ca.Di (2 .0 Di 0-rI" 0 (U 0 rP-Di 0-(2 (2~~~~~~~~~~~C D,

P-to ~~~~~ (1~~~ Di P-to (U (0- 0-~~~~~~~~~ (U o

0 Di -1 Di (2~~ am9-0. (U - '- (U P-i -.. Di 0.000o 0-'o - Di CD- (U 0

Di CP-FV- I - Di (UCCDi (U I S D

Di (U b-' C-

C- CD trz C,~D (U D. U ( 0 gm- Itw M T w C. M D.- (UC0-CA

C,* t = P- f (UC Di (Uo nD -C.00 co~P n- a = (I" D

a C. 9w I- W 0 M 0 M 0l CD2 co 'Di 0

D. i C-- = CI - -- Di"O 0L t Di CD t D 00 Di 'O (

0 - '0 ri"~~~ 0- 2 co 0 9 iP~C2(2 0 0 i-D 2 G 0-coi I.- I

- a , o - rI -i 0 - D co f ( D 0> 0t CD SU 0 C Di f-- -C> If-i ( w ~ 10 (U - C,. ..

n " -0. 0. -Di 0. Di 0-' 0 I CM. C , o. 1c.2 co C. w 0 w( 0 = (U ~o Di(U( 0 ( I- U ' -0n00(2 0- , a. -- ~ I0 -0' 0 (U; rI "9 0 -- Di tP- - (U Din 0 D U 2=p ~ Dco- ' Di 0 2 o. P-i t1 (U cm* CD CDo (U i i

Di co i 0 - U 0 D (U r .( 0 0DiPI D - > -. C

a ' CD( to to-i~- 0 U 0-0 iD

0 -to .-gI"

0

Ze,4 --

9 b4

I9 =4

I-- M

o. -

Page 140: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

Ca a W I, ~~~~b Ca a- .Ca = -

_ t> C ~ ~ ~pD = * DD

M Ca~~~~~~~0

'a -> :-

C . Ca A.o CD '.0=i

eL o - X o_ -. r' CD -~ a CD

m -. O O C C ICD Ca '¶ O '." IIC --l O a0.n t- - on u _ CD CD C_M " .

et to~~~~~~~~~~~~~~-CD * r C t4 O_ 0 a- -g w:O

.- t , , ( ..=CD Cb CD -.am D

,'t oC o, o.c o = I

a-t ao i- a vIa-~. Ca Iaa- - _* - _0D -. = C C _. . C I

- 0 Ca -

~~~~ ~ I-- ,- , -C v* O t"_

~..~~~r ~ ..,- ~.~- -. o" - o ,- cM IC C 0

CD Ca_ '- Ca t1 I L 1CD r-,- ~~~~~~~~~~~~CD0 D I

i,-,. ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~~~~~a_. = o a- - ' 0 -a C II~ ~ ~ ~ ~ a- a -~ in. a o I

cr- a to C f 9 _ W

0. ~ ~ o -_ CD, M: CD* M.CD ' 0 Ca _ _~ 110

CM M' Ca MCC> x CD Ca n e. _* cc laCr to w P v: e M. am C c .p C . O,_c

:= u W CaX~~~~~~~~( caCD I- rs02 eb 0 = C la C CD CD e ='D 2 *, M u k M CD Un_

"M M - = I. V" t o * 0 _M -CD CC l 'a. , _- e oC" > X~~~~~~a o h -.euaX"

Vto t - a aC HD M to f D D

- . Z. -* M n 0 t " ,I- OC_ CaBla~~~~~~I) Ca tDt .- C D C L W b a, i -o CDo- C e , - - 1 o -

D D tOD c co Mt Warb . t- CD U C, CD 0 C ICV11D Xb._ C a -o M M C aD a c4- I

e1 b t=t D= O * "- 1 C, to - W m C O2li.0~' * 0' C -_D CaD O- CN I- *0 0) I-D1 a- CD a- a- a :a Oa a a- o a CD CD C0 co W _ i a 0) bD C' Ca 01 c '- n cz > n a CD ~ X a -*rx -9tC CD c

11 _ fD : - cr 48 ° ~~~~~~~ txt r coM co . I>ttCt111 0 n CD a- t - CI t tn O4 '. O _-.. 0 C -9 a - Ial (2 -'- h.-C* e-C o C CD a- aCD= ~Ca I-t'D D * Cl a- a CD CD '- a C

a- o '-in Ya- _ a- I- a-'.0 (2 Q Ca a- Z ao - t-Ca Ca CD 0 a- ~~~~~~~~~ ..0gm CD (-C 11 . ::r ~' CD CD .2 0 ao a-- 11a- ~in a- Ca 'a C c|| '-o O-.r2 .- '-, a-a a CD CD Q a _ | C W

11 C C CD.' Ca a CDD- 0 a- I- - -

r-'- 0 '-in am in-'- a-. O pin-C 0 .'-n 0 VC CD- - '11~~ ~ ~ _D a- (2 I~' atD0 D aD ci - 1

1 o O cn CD a) .0 . o 11I0-.ci{1 0 fD C b O * a-a CD I

i o 0r o_|l ~ ~ ~ ~ ~ 0 a- 0) I. o e~ I- otat 1

{1 ~ ~ ~ ~ ~ ~ r Ca -9 Ca -. CD Ca D Ca-11 - c en s~~ o ta I .C I

{~~~~~~~~~51 '-i a- * l_11 ~ ~ ~ ~ ~ C a CX- o -{ C t

I~~~~~~~~~~~n' a- 0 ' Ca (O

A~~~ ~ C-' 0 I- -. Ib

-9

toa-X. Is

- c

oM 4

.0.= 9to n

CA"

.--o

ee3C-

am ICXI In-4a-4

am

-90aM v

X Cl

l�_��_i _Illli__l___l^_X__I_�I___�_��_.. ._�.�_

Page 141: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

, '1 0 0 'e (2

·w ft - ("- 0 "D

r-. .-. ( M -

t:~~~~,-,,,r , ! Da~"~~~~~ ~ C,- . I "1 -' C W-' "~ '

I r C' - C', C.- Is _-- ''

, -, n , , -, ,. la-'(2 C , W C/' ( ~~ ¢'!, ~IZ' I' ,.t C--CD C

COO f " f f f - f'- -0 0- W-¶ MO C - P C I - . CD (2 0 Co~'0 0-) * 0'f O f,

-t 0 I--..M

E I I 1:-

o ,.c , o0 ,r--,- m r.-',- 0-,- ft-.~-~ m- c-,- , .. . 3 C, ' co-'~ 0 - 0- ~-~. I-d- ¢0~. . C C" --(. -. a. U ftC",

... ,-,-o ,-, ,o .-. ,-,...< .... ,-,- ,-. . . = ,-,.co,%a,-..co 0 f. W f0 -'9 W co IC:) sof ft ft ( ft Al (2 ' 0) 0- ( . ' t-0

e -f0" ft I ' f01= - t - CD 0 .A9 - C f .0 C t f -

0-00~~~~~~~~~~ ,-. i 00 CO~0-) sa'm co W~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ C ft c. CD t ft 0 Ca- I=. - (2 M ft 0 a mU '- (

CID CD I1 ~ " S . cneg~I.,- ~

I~, ¢. , -0 0 gr* (:2 ft ' - 0 (2 .¢.~ ,.9 fco CD C 9 1 h0 5 '0 - t ( ~ (~s t.co W~C 0t 0'-a, 0 t - -.- -f

~aCDC 0 I-- - D- . a- C ¢ -I -. 0 0"-'1 1 =11 9 ft CD C- ( -" ft 0 0 a a -," 0~v C n . t=C ft 0 a-a Po 0-.0' coM L C- co I- - 05 CD f

Ca-a-a-'0 0 0 0 00 K--,(25-I 0- a-a- ft CO ft I-in ft 0 a-wa' CD 0 -C

. f - co 0

ft n0) 0- 0- C.a-a-'9 a' ft (2 CO CO ~~Co * I- 0 f I ' a- f

(2 ~~ CO ftOc C- " a' - t ICO ' S.-..C ft M9 ( 0

'v 0-0- M C t=ft ,-I C

0- 0- ft C> !a90 C t O~ Cft a-'- a~~~- -- a '- Ma(2co f

0 co C 2 C 0 -a-CA ft co a-0-eba c CO C, co a- QJ 0 co0 C

9=f e a- -ft =- -a D o 4 0 0- a- ft 0- 0)0 =L. 2 f CD '9, M CD

a--a naaa 0D 0 a-a-.=alb

0D -t W ft 0- a0 C-, C', I'4IC

A-. 2 2 C M (2o a- > 0 'ft ft 'a ft of COO ftC 4 . I C

co =O ft C, -

'0'9 0- 0- ~C', 10 0 M c, C'. 0-I0 ~0 0 - - I' U .- I'l

(2 0 0 0) 0~- a-, C- t - ai f - *a~a . b-a",'a.

ft ft U .~~~~~~~ 0 CO a ~~~~~~ U 0 CD ' 2 - 0 0 fCO a-a- 0 '9a~~ - t'e a-I---~~ CO a-. - a-a-.a. 0 -' I-

-i I- ft 0 Aa-I- -a ft w D.M CO a-I 0 - 0 -( .C. 0- a-- .M 0 , "0c ft, 0 cof 0 a It

'0 ft 0 a-~~~~ M v- MO f 0 -' fta. I- (2 M

C. 0) CO 0aI f tf 0 ft f

0) a' a.0 CO 0 0 0 a-* CO 0~~~~~~~~~~~~~~~~~~ M o C a ft * 90 CO Oft~~lm 1 '9' CO(2C ( 0 -a-IC ) O

f -0)CO C f - a- C ) 0-et, - ft ft -f'9 CO C (2 CDI 0-f (2. n CDaC,-" 0- M CO

M9 0- ft 0- C. ft

- M -- t=- E CD, M C,~0 M

0) ft 0) 20 00 C>a- ft 0C.-2 '0 ) 0

CD ~ -:1 " CD-I ftCD 0 tto a -a-a - -

"--g

-

-20we

00

I1.4e.

W C-1

0

U2

,-20=D

C2 C(2

.0-g

M =

UI2 o

0cnm

III

Page 142: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

0 Ca

#~~~ en.4 '-a~~~~~~~cr =a'

· L .e,-~1'.1 t' .-f. a'

~-~~~~~~ C C- Ca M-laQ C~~~~~~~~~~~~~~~~~~~~~i... 1~

i~ -... Icr, 10a e~~~ C. 0 rI- U3~~~~~~ C. i Ca

en / a' - " a' .-. , ~ g- . ~ a ' .-.,C, e n CM ,en in C, Ca' i ',

a' o n -- n co e n aa' - a' en c- -, ' z-* en en Mn en W

f- - eV - C CD Ma'

0 ~~~~~~ a~~~-. cm- -I

1110 CD a, Ca -W-Iin Q f-~

_C2 -, 5C ='v -0 Q =V -M -rl en a ' 0 ene '

I- C fI-M1 --aame C' ,n W .4n0 4 -~n a

M CT ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ C

( 0 .n4 in e sIC e =- a . ~-c n -e

CDe Cs-- Ce , en a am M en Mns- -

- - ~ ~ a' enient0 - 0

.4 * C a' - ~ in ~ - C>a C- n co.

~~, ~ en ~ , ML. n- ~ ia' g-~- ~- .o ~ M s-I-a =- to en-

n~~ en ~~~4 C a' 0 0 * 0 in a--.4~~c" eo o a' en ~ ~ co en g-c av"r

in ~-' t .4 en e (. en to c.en en c~ a o.C~ in in i n - en en. - I ors--i.in a a-- -- a L 0 . co c., toe i

in in .C . to

en en en en ~ M n

en f--. m -- e*D en c-e .- a'-a-

a-' en o a- a' pv- coIa~a~ en

��---·-·-�--x�--~-I----�~-'���I�'�`��--� ��--

Page 143: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

0 ra

--. (2I ,

C. am eb t-.. ~ Ca 0vi

(D, co - iCD coCa c-iCD SD.C- 0 co 0 me Ca Ca

a , C--CD 0- co a - Sa C - (2 MCa U2 (2 CD Ca Ca -... ' -. C Ca- C- c ,. CDn- Zv a, c-a, -CD 0-.. col 0CA -i CD 0 CDi

CD 0- C12 =0 C 00 a a-i, 0 Ca 0 CaC"- co -. CD , ,- .-

c-o Z cmci i'0 -0 c--a hC

Ca . Waa. rn ciC C -iCa 0- eV- ED 0 ca D "0 C,., 9WC c-- co C C~0 -

aCa C b P--M a a-two cao CA0 72 - Pl --" 2C Ca a--o-0 C.0- -C - Ca CD el c-iD 0M4 CD CD L

CD r-CVu M-- CD 0 Cam r - C'

C I C I '. ." CD toLO a-i 0-0 a-~ ~ ~~~~~ c--a, cu a- a0 a- a-CCD M C-- v-- C~ D CD CD 0 CD - C a - 1CD 0- ~~~a CD CD 0 Ca, tva- CD' ~Ca 0 -a i-~ 0- a C.i - -a. CD .Ca .

0- - CD 0.4 a-0 CD -0-rnO~~~~~~~~~~~~~~~~~~~~~~C CaC .0 ii 0- Ca ~~~~CD l=2 --

co I. P - c .0 c-i- a Ca 0- aW- l C Ca0 c m a. CD - .= co DCa 0 0

a-i ~ ~ ~ 0-I CD aa II 0 0 a-"C"'CaCI- C12 CM K~ Ca- j=--- 0 ' " .- 0 = , 0* ci Ca CD a-"M

~C~~~~ ~ ~~ ~~~ ,.l -~- "' CD (.2 C C-i- M D* I C: --. II~. x : I -I

,~~~~~~- c-,- Ca. Ca Ca Ca0 CD 0.400 II~~~~~~~~~~~~~~~~~~~~~~~~~M CA0

Ca CD 00-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~v c

,", - CD e', = o -

CD a-" (2.40 a--s Ca~~~~~~~~~~~~~~~~~~~~~IM a

C 2 a-W-a-

0 ~~~~~ ~~~ ~ ~ ~~-.o ci 0.-W NM M CD Ca c-iW 0 -CD.Cv, I= :1 -a --0 C W -

.Ca ~ 0-a-C .a,-

ad . I--'- 0 0 C -- CD a --coCD a, 0-- C a-a-- 0 Ca U a--i~~~M c C-0 eu e--Ca 2 - aa a 0 1W( m En 0- 0 CD-CD c-- o0r c-Q -( C L-4c-a CD --CD a-- .0 a Ca - -v a- -n o = W n a-*= CA0a-OcCa 0M(2C .4 - CaC 0C.C -i-ca, Co a.

0 a acoa , CD a-aCC - Ca cD Ca (2W (2 C C

CD - 0 (2 a-" a-i 0- 0- CD Ca a-~~~~~~~~~~~~~~~~~t - Ca, a-00-0C. - a-C aCa 0 D C Ca- a-eC l a--cc 1(

~~(2 Ca a-i c-i- CD Ca Ca c-i-Ca "0~~~~~~~~~~~C M--c-- C- CD C12 c-i- (2 o C 0 a ( D 0 -a0 C

O0 CD CaDC C> M. M M. aa i a Ca i

M 0cD CD 0o - a ,~C IVCDm0- M W-a- 0 CD a-- CD 0-co a- - ca,-' CD a-v = .0 - 0- 0 a,- a-D - 0-

CD Ca P-9i~~~~~ 0 a- Ca 0--~~~C--0 CDl a 11 c-i-C 0-6 1=Ca1 a- WC Ca a- Ca W- 0 C Ca CD c- C C CD cD CaW -

2Ca 0(200- c-i-CaC- a-co (2 a- o Ca , 2C0 0 (2 b- . to a-- CDCa D aCD C-(

P-i-a(2a-0a-0 CD M M( M =r 0D 0-i CD. C.---C c-- I 00W C a I C a .0 - v-i(l 0 a- S.-- 0 ~ C 7 -. 0 0 = 0De

CD C.-i- a-t7 - a-CD 0--2 -C> - C0- . -i cui a-i ar- (2 Ca (2 0 a,-" a-. C -i- a-i - Ca ca M P-l Ca Co-a-' -a

Ca 0- 0 a- a.C C-- 0 0 C 0 0- 0 a a-. D aa00- Ca a--'la . -Ca P 0 Ca Ic--CD - - Ca 0 C CD ci CD CD 0 Ca CaM CDC>rW CD C 0 C M v- C> 0 -

a-i Ca -i- CD 2 0 Ca. CCa= co-i 0- 0- c P-'ato a a C a-2. WC Ca0 C - a. -- C C a a- ca-o 0 : C-h a - . C c--a- -I 0 Ca, a a-- 0 0 ~ Ca 0C-- 0 - D rn C -i - v-i- --Ca Ca C (2 ai 0 -

0CCaCD C a ~oCa 0Ca -0a n- 0 CD 0 o a-i (2 0 C -- CD -a-C S.i 0 C "C CD i2t CD aa =C> il 0Z0-0 v --iaC a P-' CC- 4 .4 00 Or Ca CD 0b ~ 0D co D . ,Ca m v---iC CD - " 0-CD C> v- t Cao

C0 a-i C a-m.-. - 0 0m 0-C --a- C ii -- CD - a--CD (2 i

co 0 0-r a--, C Ca Ca C -0 a-D a.( a-00- 0 a-aa

c-i C c-- a C a - C a C CDT CaI CD.Ca va - -C (2 ca

00- a D a D - - 0 CIC Ca C-i . a -

Page 144: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

:m ~-*---,~0 a, o o I -,, W co 0 ' f (' t' 0

W 0 1 rb C- 0S

a3 W 0 -.

Ca r , C

col i 'll . QCo : C M C

OW Cb 0e C --. 0 -

b 0 0 0 1 0 "C' 0 o -- . 0 2 o ao a , ..

Ca .~~~~~ 0~ 01 t a F I U2 -0 SuC co

aC a0 t W0 t CD0 0 W C!

0~ M, f '" O:: i *-~ ~ -J C 0 E:Q rD0k

DJ W E~t =W~ - -, - -oJ C, o -. , 0 -MOcd* Eli ~.-.'- k.-'- 0 ~ 0 I. -C -,-9' M'J M.

010 a C- M' U .... to " 010C Ca

o a r-4'--- a, * 0 01"0 o-., ~-.-,0 Ca . ~~~~~ Ca U 0 0-~~~92 0.. 01 OJC. 0eV- 0 - 0 .- a

0 IU ~t, 0,. 0 0 ,0..

r* ~0 0 ~ Ca q0 II C::a 1-4- " U .0aL C - C" W- 01 0 f 0 01 .CW

:,=Q ' - -, = M: 3"

t= lai - W 0- W C, 0- 4 ' 1W CD Co W 0 Z

01 0q. 0 -( 1 CD ato- 0 .0W - 0 .

,m ~ p , C "a iza t7 C -0 ~ 0 M I

, .U VW0- - 0 (D0 0--C

IIWa 0 W C. :, CO

,~ 0~~~ 0 0-0 ~~~~~~ O U 0 a, -~~~~~~~~~-W 0 IL0- 0, Ca a, 0 ~~~~~~~~~~~- '1 0~~( M-4 C- 0 1

k a-. I0 1f k.,' <~ C~' -- 'f ~--P' a-.. c0* Su 0 G a4 J I I-ICO~~~~~~~~ ·6 M: =~, Cq" ¢"'f

to 0a CWa -CI-,a IC 0 C- I CD cc, O C- c1-4- a, Ca U~~~~~~~~~~~C t 12

Q o o- f-f0 M 0 r4J 0 0 . 00 C0 o W-a0 1 aa to t 0 C. 411.1 aC,

too a ~ CZ e-" co1 .. '.' e a ',Atoi cu CD0 ai- oe - C

L0 -. 0 ''V aW a, Ca a--

co (D co a, W W~ . - . a : , 0 M

C> M W 1w -W4 to -D C- (D 0t, G*

c ii ii i S.oc i'lllll~ e-Q -C ~~C-, - (a Su-0 a' . lC-, Z Mcgm , to 0CU" c -* 0 no0

0 0 '~~~~~~~~~~C "V , Ca , a oCZ co -- 'E:a*0* - 0

to t, 0, 0 C-, ~to a , -C

V.= cu C- r- - CiW ' 0 - Al 1 Wa aW -K. " rb0 0 0 0 0 a . a- 0

fu . co a, "0 a, Al0M0 - a M,CD rb C2 CD~~0. W-- a-0 - a,

tv 0, u 0 PV~~~~~0 -0a-

I CD, ~ a -a.0 a 0 , C--C -C m

Ca W 01 CaC.g , Ma--a n -- a C

a, Ca~~~~~~~-o,-o11 0 a, ICa ~~~~~~~~~r -- t I coj ___ ____ ____ ____ ____ ____ ____ ____ _____ ____ ____ ____ ____ ___ ____ ____ _____ ____ ____ ____ ____ ____ ____to__ ____

(3

CI.4

=- a-4

CJ'.

a-.-.

co W'mam.4

rim

De

�_��_�____1^1_______· �__�__·_1_1__�______���

Page 145: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

III

,- a, Mc~~i c,~~~~~m 2 c w

i, U-..r ..

c o. 0C- O

Co~ ~~~ ~~~ ~~~~~~~ ~~~~~~~~~~~~~~ CA C , C -',-. t$, -. o , C:=~ I{~, .

co~4 ~'- 0 ' c-' ' L. c

* 0 (~0 ~Di ,..2 ~ 0 (b-j

(' ~ ~ c 2 Di ~ 0.. ( eAl ri ..- b .,

CDP

.n I- = C(D 0~~~~~~~~~~~~~~~~ -1 ~ ~ ~ ~ ~ ~ ~ ~ , .- .. Di0 0-. 0 -- 0 * (2 ~~~~~~~~~~~i Di ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ -~..a D P

co, co 0 0.. -~e aD . CD O- -

En CD 10 CPM C9 E M

CD 0. -~0 0 (2 i b

U~ ~ ~ .- U ~0 mmm ~ 0 r.'*Pf

~"", ~q , -,- Di Di Di 'I - CD Di CD ~0 0

IDad C-D p - ' En ~ - 1 (2, 0 P~ 0 ~""-""I-" (2 10 i M ap~p

o 0..Di eV D I iCDE

Di CD I~~~C

(2 ;IC 0 0.. =

M~ -* CY, e-q CD CD 1PaD

eff C a, (9 C= o-- w M U >w C C

, Di 1 1 D

I= C. UD 0ID ID n n o- M C

CD ( - . M I--i C . la,En C> CDC, 0 C CD

r~~ PD CD -CD g CC, M-*, D

En 0 0.. '1 0~~n Cw S n pDCD cn.

o..~~~~~~~~~~~~~~~~~~~~~~.

to M cr I- 0M --~CD ~~~~~ Di 0 0 0.-I~~~~~~CM

C~JCoU

-'4

Page 146: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

e,- , -~ CO cM C

.~~~~~ ~ C"9' 0 .,~{~,r"-~-~ . ~-

.~~~~~~~~~~m-. , 4

4'm ~..-, r9 0-, CQCl{t,'C,

I' 0~~~~~

I ~ ~ ~ ~ ~ ~ ~ ~ I~C,~ ~ ~ ~~~~~~~~~~~~~~~~~Y~~~~~~~ ii-0 -'. M M

I I~ ~ ~ ~~~~~~~~~~~It -n

00~~~~~~~~~~~~~~~~{~-.. ~ ~ ~ , aC, Co ~ ~ ~ ~ ~ mh

'"' ' = '- r '- IIinco i 07 P-M S- n co~~~~~~~~~~- 0m c

Co o as .

C> ~ ~ ~ ~ ~ ~ C

C.am

-. rE -C -

W CC C -cIV V 0 -.M

M--5-4 C, 0 C a MiC

hi C, . 0 " C.5-10.C a, = -o -I

e~~ 0 5-' S- a 5

h '-~~ '-p e- 0- C 6 n S= - I,* Ca w - W t. =, =-f

I I 5--. hi C, ~~- t

W el =~ W n c h, MI I C) 5 'W -.~~~~~C> 1 D

cr 0. z Q,. C' "a - Ctl-~ a- PC

10C, 0 l

Ca C-

C* C2 cx C> ~ ~ ~ i , (--~0t02 M C- CL. co W ~ ~ '0 a , *' C

C 00--0 0 ~Yr-CP C. C. - : M-5' a C ,i

C,. hi'V ,

CD "I C. ~ 1I 4

to: I.w IC, C6~~~~~~~~~~~~~~~~~~~~~~~C

Mc C- hico

-4

-,

IS--I

-0tomml

Ca

tocw

!W1wUS.I:. A

tw I.-420C, MW

31=L-

4w03

Page 147: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

"a:, I,.- ~- (' (I .~~~~~~ ~ ~ ~~~~~~~ ~~~~~~~~- =-~ L-

,co C,-

i -.

QJ 1-4 0.. -.~~A n =. %71 0 1-1- CD C

i-. - (C '~ ~,. ~,- C~2 (c M '9-.. w ' ,- 0 I1 I F- co-. 0.. - C. o- .: -, ' : " co "o ,. me

mmm ~ ~ ~ ,-,- ~, ~~~~~~M~ o M=r . M C Al -~ C, '-'.0.. a, - c _ - aaCD " C C ID C C>0 0 ruC '= (nCa '9 Ci M, -. C n C-CD - .-, , o c --. - = ,-- .a 0 a, 0.. =- oCCD 0. C a Ca 0 o C= 9 Ca 0 .Ca = w a, CC a, - toa '

a to to 0 ,- -0l~ a, o..cn a, , 0 a-i - , -a co1a,

o'" fD.) C~a-~ =C ¢-',- C oC, ,-- 0c

0 =-, .0 == 00 a ' -CC m a CD , -. 0 aa o- to a-- CD' a, co -CD- CDC ~ 9 -~ CDm Co- MEi gn -a

'9 Ca co ..~a -~ c t-!-'0 0~~~~~CA- CC Pr w nC

C> C wfOa 'CD a.0 , to0 a, t C r - mm a,

I I~~~~~~~O I am0~ a, 0l~ ~ ~k-..-- Ca, -I- ' r4 0~~~~~~~~~~~~~~~~~~ C>. r-

a, (3 0 a, Cd . - 0 CCCC 0 a , n "ar 0. a* ii M L.1-0m- 00 '9 aD 0 a, 0!-0 0 0 bc~~ a, 000...- '9 ala tv0-0 to1

M co CA.. C,~~~~~~~~~~0.0 0 CD a, CD- - ' a,

0- 0 a-~~~~ a-. 0-~~ 0 a- -' '

CC 0 Ca Ca-CC a~~r, coC", 000 Ed

0r C. 0. . Ca C- CD Ps

0 Cd a-,. * a-* C- 0 - a, a, 0 a-*4-a- 0 0 a, a-.t~ 0 -. a, -* 0-0

C, 0C.4a , 1* ' a C (-~~ 0 0 a-- 0- CC~~ rn a, 0 0- 1 CCC a, 0 a, ~~~~~c o 4a' CC - a,'9o*

a- rn0 Ca* a C> C9.0-CD

CC '-'. o -g a,~~~~~ M. 0. . I M 0 '

* ar 0 , . CQ CD co

a,

'9C. M C) Cox*'.~~0 Ca C-

0CC C . CM. CD0a-' 0 -a '9,

o a'. (C 0 ~~~~c PO- -- -U 0 r a , - DC-C.01-' a, a, CDa 0, a-' 0 -. -

rn l--a co0 ra0 Ca 0Ca ' la ' " - Al- 0 a, -(-a-a--.eb - " 0r

C-2 -.~~Ca a- , a , 0- w-5'9C 14 '9 CO - -= C-

0- ~~~~~~~ 0 0 a,~~~~~~~~~ '9 (C. to *'9 C>a , -- 'm a- -9 . . ~. ' C C

0 0-c--I- (CO a-- a~~~~~c C7~, gw- '9 BP~~~~~~~~~6 co M C,0 C0M= - 0b , 0 G~. a 0

III

Page 148: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

~~~~ 0 ~~~~~~~~~ 00c CD nim. n~ ~ ~~~~~~-

oob CD -w

r1 ~~~~~~~J 0~~~M- = C, M- CDr ~

eno -Cc,.f, co , = d. -

C2 3 tm - - Vt cn "' Cos

0 0 t* t= co - Cfl U3 I= 'd

~~~~~~~~-o- ,'oQJ ~~~~~~~~ en r~~~~~~~~~~ 0 B' -~~~~~~~~~~~ B, en '- M

L~~~~~rb I~~~~~~= a -4 -w -'CD..* "'=r D' 1 1--I WM/ to J ~

et,, ' B, *W

- 19- - en --- -- 04 0~to aC, - ,--

* (.-i- =

0-~~~~~~~~~~~~~~~~~~~~c 0 0en~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~e

k..~~~~~~~~~~~~~~~~~~~~~~~~-I

t= = I~~~~~~~~~~~~~~~~~~~· -, c,= = =.gu~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~D CIC0.0~~~~~

~~~~l ~a, ~ ~ ~ ~ ~ ~ ~~~B

~~~~~~~~~~~~~~~~~~~~~~~r- ~ i

I-'.~~~~~~~~~~~~~~~~ I- ' I:-I"

-I-.. ' .- . U0 C-I- w-'B"~~~~~~~c C,2

m. a e cs= * B, - -~~~~~~~~~~~~~~~~~~~ lb~~-

C>"~~ ~ ~ ~~~~~ C3. 1 1~ (2 en~~~~~~ ~ ( Bo

lb~~~~~~~~ ~ ~ i I- I~rb~ ~~l - 0enco0

- -~en n ~ en ~B, B' en(2 - U - ~ n B

~~~ -. ~~~~~~~~ -~~~ M a -- B, n -I I- l -' 0-tb --(2 _ lb C~~ lb h.'- ~~ lb C--a C-I

Cb M. - .'a C- - w 0 co n 1 B *.n

Mb I-. c, B b - , ~ B' ~ e o lb -Me

'-we lb '-enB C2 en b-l C- - 00 - C,M w'~ to0-I- SWI - Su C - M w I _ '

L- 2 , C 02 O CD-r- co~ 40 ~~~~~~~' 0- lb~~~~~~~~~~~~~M 0- l lb -B,-en B Cb MbB,

e-'- 0-- =.~~~~~~~~~~~~o C.-t M2( 0- lb - e B, - C:> t '- r-- 0- - 0l

C= C> - - 9' I"- 1 Clb CD

lb -g t~~~~~~~~~~ B'~~~~~ 0- lb~~~~~~ !a co - . v B n I-

*o Mb 1B ~CI'~ t-w- w

Page 149: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

111

n ~~~~I CO g0| su I D

o '7, = '~ i

', - ' ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ -' U m a :n O n n ::s O C rs t~~~~~~~~V t= M

C~~~~~~~~~~~~~~c on cC> _ > S ._

0E~ ~~ I 0

en e~~~~~~~~n 411 .. I . ~ .) C . - 11-~,C

~~~~~~~~, " W CD C-n

IO t M .I o , 0J -X 0' CO t o - C O

e n en 0 _'

u| "1C C~ ,.,,:,.o b w~

'~~~ ~ ~ ~ ~~~ ~-''s C a)~~~~~~~~~~~~~~~~~~~co co :O.l~~~r =D = I Cr

0 C. _w

CDw _

C1n . C D c

I--( -il ~~~~~~~~~~~~~~> °M fli Ic o.T. _ C>

0c w

C>~~~~~~~~~~~~~~~~~~~~~~I

II 04 1 M c0 C a$ o11~~~~~~~~~~~e I-' I ni{I I I , H Cj t pil~~~~~~~~~~~~~~n I IIo,1

1& I . i . . z ... .11 o I = I ~~ 51. 1CDC

,, , Dp~~~e - ObI- q CO Co n I :s I t-n o C-1

-~~~~~~~~~~~~~~~~~~~~~~~~~~~C1,-f Co , _ _ _ _ _ _ _ L7, 1 . _ __ ,

-C cc xe o C.) l

- - = C) .i |a Q

92I 1 I= !i cD .0 c - C, O ., c t, co 1

|| C. M toU >{WD|>1ll tl b I : " I fCD I n

il - os | ~~~~~~ | D er 1 ~~~0 C >11 tt> I O I _ _ I O C I I~~~~~~~~~~~~~~~~~~~~~~C

11 n I O I _ _ I - l = I I~~~~~~~~~~~~~~~~~~~~~~e

..a -

o.0 011 or

0

en.-to-I

Page 150: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

c .,o k=,,-

:f C-- cm.-

O ~ ~ ~ ~ ~~~~~~~M1 C3 O~~~~~~~~~~~~~C

=='o ~ ~ ~ ~ ~~~~~~~~t

CD ~~~~~~~~~-. = ~

k,.,,~ ~ ~ ~ ~ ~ ~ ~~~~1 M .=

V~~~~~~~~~~~~~~~~~- cm ~,*"

C, . .- cm. ,IV~~~~~ ~ ~ ~ == ,-,=~.,.~~~~~~~~~~~~~

I ¢='~' 'q~ ~ ~ ~' i~"~ i~.. !.,.'

~~~~M

O w C-. ., ll.*>~

. ~=,~ ~ ~ ~ ~ ~ ~ ~~~~~~~~~~~~~~~ ,. -,. 'n,-

V~~~~~~~~~~~~~~~-..

•~~~~~~~~~~~~~~~~~~~~~ M.,=~ ~ ~ ~ ~ ~~~~~~~c C= CZ,-

~,~~~~~~~~~~~~~~~~~~~"·- to =1 C. C. .-. =

r=, .. CD

III I ~ ~~ O CD 0

Page 151: AN ARCHITECTURAL COMPARISON OF CONTEMPORARY … · aggregate, and to update information maintained in multiple, distributed, heterogeneous systems. A typical distributed heterogeneous

IbHS~~~~~~~~~~~~~~~~~ pt;S l~~~~III

L-or rO .

C-. 0 0 0) 0 CI 0 C-m 0-C-9 ow --0 I0OW E.

o ea C,,, · , -Co CD- Ca e

C"9-

CD

0) C-I- -- C-, 0 II . cc I to Co D oC- a D cD ~ o - b4 .CD CD - o- o-C o. o -Co CD . v,-

0 CD CD(OMD ~ CI -'U CD CD CD C- - I- M3 00)0)C.

CI-U 0-~ 0) 0 ~-tC- "- CD "I ,- C--%o0 0 C o CI- 0 D 0 - I CD U W r W CD C-I 0DoC CS, .'C CD 0- - -I 3 =

0 b-i., Al C ) ty :'

0) CDIi 0) = a,_ 0-- -- .CI CO- CI 0 C- e-~~CD~0 Co Co- 0 (DO) ..-'0ii I~ 0 0::'" .

'-0 M CD 0) 0) 0 W CD 0; - 0I C * 0 C0> 0 0-0) CD 0) CaC-I- I-I Co b-" C " C- C---

n cc (D C~~,c'CD)Q C > C"- C M h --I' 0-.C 'aC Co -I 0 ' II"C-, .C S= -0 -- 0- CIO D 0 0 0a- C ( 0 Co 0

CD 0- 0 C-I- 0 CD', CDI- C D 0- 0 C * I- '' CD

_ 0 ) 0 C D 0 C C - tD 0

Ca"'- O W P-rW b...M0C 0 - CD 0)00 0 = 'oa(DO) C > 0' -C -iC CD Co C 6 -I 0Ii CD CD ~

-I n -i <D CD O. -, oo 9-a C..C - 0) C. Co CD C 00o C$- . - Co 0 C D n , - 0C- CD C> D 0) C

UCD am cc i,0 C 0 00 :) to 0 CICD 00t C MCD o C S= C- I .0 Cod CD t 0)? Co - 0 0-C

0- , t * A, o) Co W 0 I-? 0 a -- 0 C- CDI 0 C

0 O'0 0) I C-I-0 Co I- ,- Co C-I- '1 '-'V C --

IC -C -I- S .- - C ,- 0 -to - I CDCD CD

C D1~ c o~ N C ~ m " I- -- <22} ,- 'CD' 0) CD 0 ) C

CD C -I- 0 -C D - c

,-~ W co ' 'a0- C CD 0

CDI ~~~~~~~~ ~~~CD ~ -.~ CI M-

0 - Co C-" , M

CD CD v- =-I C- Co

0 0 CD MC- ~C+, co C- C- ~ ~ ~ ~ ~ ~ C-- -C0 C- -V.-3 C~~~C -I ' -I

· -. . C W : , = - C O M- CCD 10 =1 - CD C> " "0 CD C-I 9- e- C-I-

~x. '~~~~~~-I I

O nC - C

to

CCDD

l bC- I - 0 C o O I - ' ~ C - I --9 C Y , (1-

,-. C. a0 - 0 *?-'.0 0o t C )0 0CD ~0::m, -C D o-I- *~ , 0 ,- -. % C al~o CD 0 Co. C 0 '0 0 0 - C

00--a~ ~~~~~~~~ ~~~~~~~~~~~~~~~O --- 0::m 0 -f -I0 II'-0 0--i. 0 ) 0 CD (~~~~~ C C 0)0 M CD I

I C0 ) CD 0. CoC--I -- ) 0

0-h- ) .'.0-..C CD~o C. - tI0 C CD co C -W CD CD CDa, 0)I a,~ CoO I 0 -C r

- C-I-0 Co~~~! CCI- D 00) 0) -I

CD 0 0 CDC C o"

1-b C O 1. 0 U 0 ) 0 0 CD CD

Al co I-.- -C- 1coco ' C U U -P CD 0 -C C. D0)0 C CDc, C2 I- ~ i 0 C , C- --9. C = v- - 0 Ico W C-' C- 0 ) CD CD

co 8 0 1 co C> 0C- C.CD C-I C

0 8 0 = - C, - 8 ) OC-c~- co 0- 1I- - C0 I0 -- -O C .II-- C- -, co

0) C - M 0 M V 0 ' I-CDI- CD I . C 0= 0 e L. C ) 0 o C-D C

CD a, to = 0 . -C"a 0 ' 0 0 .C 0 0C C CD{M "' '" 0)eo C-C C D 0 ~ ~.-C- . -. C, -- a, C I C-, I

0 4 C--0- Co ~ C"- 1 C ~ ~ U --- 0 -

Co ,CD> 0-'- 0 "1 CD I- " . 0 . b '0 C- I- ('b c , 0 .CD .

0 0 0 Co. -- I-0 0 C D-" CD M la .

('1)- D D D C 0-. C-I- Co .0 ) D "V - 0

, , : -0 8- .-4

- Co U -"O0 0' 0I - ;-. 0 I ~ 8 -' Co 0 '-0 -I-0) 0 C.....~Co { 0 0 IV 0 -- 0 0 ) 0 U 0

C 0 0 CD C .) 8.0-0 CD 0 Ci-) 0

-4

C).CIA to

o ..

0 D*CD CM P

!>-C>

IT

C s DC>h

D ..

-m

cw

{,-.C)= -o

-