Introducing a Framework for Scalable Dynamic Process Discovery€¦ · Introducing a Framework for...

18
Introducing a Framework for Scalable Dynamic Process Discovery David Redlich, Wasif Gilani, Thomas Molka, Marc Drobek, Awais Rashid, Gordon Blair

Transcript of Introducing a Framework for Scalable Dynamic Process Discovery€¦ · Introducing a Framework for...

Page 1: Introducing a Framework for Scalable Dynamic Process Discovery€¦ · Introducing a Framework for Scalable Dynamic Process Discovery David Redlich, Wasif Gilani, Thomas Molka, Marc

Introducing a Framework for

Scalable Dynamic Process Discovery

David Redlich, Wasif Gilani, Thomas Molka,

Marc Drobek, Awais Rashid, Gordon Blair

Page 2: Introducing a Framework for Scalable Dynamic Process Discovery€¦ · Introducing a Framework for Scalable Dynamic Process Discovery David Redlich, Wasif Gilani, Thomas Molka, Marc

Agenda

1) Motivation

2) Scalable Dynamic Process Discovery

3) Framework Details

4) Conclusion + Demo?

2/17 Motivation - Workbench - Model-Transformation - Meta-Models - Analysis Tool - Conclusion

Page 3: Introducing a Framework for Scalable Dynamic Process Discovery€¦ · Introducing a Framework for Scalable Dynamic Process Discovery David Redlich, Wasif Gilani, Thomas Molka, Marc

Motivation

3/17 Motivation – Scalable Dynamic Process Discovery - Framework - Conclusion

Examples

• Healthcare

• Security

• Customized Production

Processes

New Challenges Emerge:

• Frequently changing

business processes

• Big Data: Many

hundreds/thousands

events per second Situation for today’s businesses:

• Globalized, highly competitive

environment

• Business processes are

“… the most valuable corporate

assets” [1]

[1] Ammon et al.: Integrating Complex Events for Collaborating and Dynamically

Changing Business Processes. ICSOC/ServiceWave 2009 Workshops. LNCS,

2010

Business Process Models @ Run-time

Page 4: Introducing a Framework for Scalable Dynamic Process Discovery€¦ · Introducing a Framework for Scalable Dynamic Process Discovery David Redlich, Wasif Gilani, Thomas Molka, Marc

Business Processes

4/17 Motivation – Scalable Dynamic Process Discovery - Framework - Conclusion

Start Event

Parallel Split

End Event

MergeDecision

Parallel Join

ExamineThoroughly

Decide

ExamineCasually

CheckTicket

MergeReinitiate Request

Pay Compensation

Decision MergeDecision

Legend

Start/EndEvent

Activity Parallel Split/Join

RegisterRequest

Reject Request

Decision/Merge

Control-Flow:

Performance, e.g. Instance Occurrence, Activity Networking Time, Probabilities

Resources, i.e. Roles and Resources

Data, e.g. associated transactional data

Page 5: Introducing a Framework for Scalable Dynamic Process Discovery€¦ · Introducing a Framework for Scalable Dynamic Process Discovery David Redlich, Wasif Gilani, Thomas Molka, Marc

Business Processes Standards

5/17 Motivation – Scalable Dynamic Process Discovery - Framework - Conclusion

A-priori Business Process Model

System in Use

Deployment/Implementation

Model Extraction

A-posteriori Business Process Model

Timeline

Graphical Standards (BPMN, UML AD)

Interchange Standards (XPDL, BPDM)

Execution Standards (BPEL)

Diagnosis Standards

(BPRI, BPQL)

[4] Ko, et al.: Business process management (BPM) standards: a survey, Business Process Management Journal, Vol. 15, pp. 744 - 791, 2009

BP

Modelling

Standards

[4]

Types of

BP Models

in Relation

to System

Page 6: Introducing a Framework for Scalable Dynamic Process Discovery€¦ · Introducing a Framework for Scalable Dynamic Process Discovery David Redlich, Wasif Gilani, Thomas Molka, Marc

Process Discovery

6/17 Motivation – Scalable Dynamic Process Discovery - Framework - Conclusion

Start Event

Parallel Split

End Event

MergeDecision

Parallel Join

ExamineThoroughly

Decide

ExamineCasually

CheckTicket

MergeReinitiate Request

Pay Compensation

Decision MergeDecision

Legend

Start/EndEvent

Activity Parallel Split/Join

RegisterRequest

Reject Request

Decision/Merge

...

-1632565513 | 95 | 96.122829 | register request | Kate | finished

-586082438 | 95 | 96.122829 | check ticket | Mike | registered

1565580184 | 95 | 96.122829 | examine casually | Tom | registered

-1413312460 | 95 | 96.122829 | check ticket | Mike | starting

-1731332071 | 95 | 96.618581 | check ticket | Mike | finished

82015948 | 95 | 96.122829 | examine casually | Tom | starting

1963329373 | 95 | 96.356787 | examine casually | Tom | finished

-192289498 | 95 | 96.618581 | decide | Boss | registered

-911496176 | 95 | 96.618581 | decide | Boss | starting

-1557314974 | 95 | 97.116592 | decide | Boss | finished

825731328 | 96 | 97.263912 | register request | Kate | starting

321550032 | 96 | 97.515445 | register request | Kate | finished

506921686 | 96 | 97.515445 | examine thoroughly | Mike | registered

721713237 | 96 | 97.515445 | check ticket | Mike | registered

-1666345498 | 95 | 97.263912 | examine casually | Tom | starting

-295525236 | 95 | 97.543538 | examine casually | Tom

31913201 | 95 | 97.263912 | check ticket

1251341738 | 95 | 97.326981 |

...

Disco

very

Discovering actual behaviour

No a-priori information

Offline: based on event logs

Process Discovery : (e0, e1, …, en) BPn

Page 7: Introducing a Framework for Scalable Dynamic Process Discovery€¦ · Introducing a Framework for Scalable Dynamic Process Discovery David Redlich, Wasif Gilani, Thomas Molka, Marc

Complex Event Processing

7/17 Motivation – Scalable Dynamic Process Discovery - Framework - Conclusion

Disco

very

Capturing, Filtering of low-level events and Aggregation to high-level information

Specializations:

◦ Business Activity Monitoring (BAM)

◦ Event-Driven Business Process Management (ED-BPM)

Enterprise

System

Event Processing Engine

Page 8: Introducing a Framework for Scalable Dynamic Process Discovery€¦ · Introducing a Framework for Scalable Dynamic Process Discovery David Redlich, Wasif Gilani, Thomas Molka, Marc

Conceptual Goal

8/17 Motivation – Scalable Dynamic Process Discovery - Framework - Conclusion

Enterprise System

System 1C

urre

nt State

of B

usin

ess P

roce

ss

Even

t Stream

Event Processing

Scalable Dynamic Process

Discovery

Reasoning

What-If

Optimization

Prediction

:System 2

System 3

Dynamic Process Discovery : (en, BPn-1) BPn

Page 9: Introducing a Framework for Scalable Dynamic Process Discovery€¦ · Introducing a Framework for Scalable Dynamic Process Discovery David Redlich, Wasif Gilani, Thomas Molka, Marc

Scalable Dynamic Process Discovery

9/17 Motivation – Scalable Dynamic Process Discovery - Framework - Conclusion

Goal: monitoring one or more BPMSs in order to

provide at any point in time a reasonably accurate

representation of the current state of the processes

deployed in the systems with regards to their

control-flow, resource, and performance perspectives

as well as the state of still open traces.

Characteristics and Requirements:

1. Extensibility

2. Detection of Change: Reflectivity; Dynamism

3. Scalability/Algorithmic Run-time

4. Generalization/Standardization

5. Accuracy (-)

Page 10: Introducing a Framework for Scalable Dynamic Process Discovery€¦ · Introducing a Framework for Scalable Dynamic Process Discovery David Redlich, Wasif Gilani, Thomas Molka, Marc

10/17 Motivation – Scalable Dynamic Process Discovery - Framework - Conclusion

BPMS 1

Run-time Processing of Standardized Events

Run-Time Event

Processing

Event Hub

Global, Standardized Events

...Events

from BPMS 3Events

from BPMS 2Events

from BPMS 1

Footprint Interpretation

Dynamic Control-flow Interpration

Enterprise System

Trace State Monitoring

Performance Footprint Update

Resource Footprint Update

ResourcePerspective

Performance Perspective

Current State of Traces

Reasoning on Current State

Dynamic Footprint

Control-FlowFootprint

ResourceFootprint

PerformanceFootprint

Dynamic Resource FP

Interpretation

Dynamic Performance

Interpretation

Control-Flow Footprint Update

Current State of Business Process

Sub-Footprint Configs.

Control-FlowPerspective

BPMS 2 BPMS n

Trace State Interpretation

Open Traces

Global, Standardized Events:

• Process ID

• Trace ID

• Process Element

• Timestamp

• Lifecycle Transition

• Resource

Event Hub:

• Extensibility: integrate new

adapter event format adapters

• Time-normalization (deal with

different time zones)

• Mapping to unique Trace ID,

Process Element

Page 11: Introducing a Framework for Scalable Dynamic Process Discovery€¦ · Introducing a Framework for Scalable Dynamic Process Discovery David Redlich, Wasif Gilani, Thomas Molka, Marc

11/17 Motivation – Scalable Dynamic Process Discovery - Framework - Conclusion

BPMS 1

Run-time Processing of Standardized Events

Run-Time Event

Processing

Event Hub

Global, Standardized Events

...Events

from BPMS 3Events

from BPMS 2Events

from BPMS 1

Footprint Interpretation

Dynamic Control-flow Interpration

Enterprise System

Trace State Monitoring

Performance Footprint Update

Resource Footprint Update

ResourcePerspective

Performance Perspective

Current State of Traces

Reasoning on Current State

Dynamic Footprint

Control-FlowFootprint

ResourceFootprint

PerformanceFootprint

Dynamic Resource FP

Interpretation

Dynamic Performance

Interpretation

Control-Flow Footprint Update

Current State of Business Process

Sub-Footprint Configs.

Control-FlowPerspective

BPMS 2 BPMS n

Trace State Interpretation

Open Traces

• Computer-oriented representation of the state (as matrix/vector)

• Size independent from number of occurred events/traces – only

number of activities and resources have influence

Enable scalable footprint update

• Footprints do not consist of absolute relations but rather relative

Dynamic Footprint

Control-Flow FP:

• Before First Appearance (M)

• Eventually Follows (M)

• Direct Neighbour (M)

Resource FP:

• Activity Association (V)

Performance FP:

• Instance Occurrence (V)

• Activity Networking Time (V)

Open traces:

• Reflective state: no relative

but absolute statements

Page 12: Introducing a Framework for Scalable Dynamic Process Discovery€¦ · Introducing a Framework for Scalable Dynamic Process Discovery David Redlich, Wasif Gilani, Thomas Molka, Marc

12/17 Motivation – Scalable Dynamic Process Discovery - Framework - Conclusion

BPMS 1

Run-time Processing of Standardized Events

Run-Time Event

Processing

Event Hub

Global, Standardized Events

...Events

from BPMS 3Events

from BPMS 2Events

from BPMS 1

Footprint Interpretation

Dynamic Control-flow Interpration

Enterprise System

Trace State Monitoring

Performance Footprint Update

Resource Footprint Update

ResourcePerspective

Performance Perspective

Current State of Traces

Reasoning on Current State

Dynamic Footprint

Control-FlowFootprint

ResourceFootprint

PerformanceFootprint

Dynamic Resource FP

Interpretation

Dynamic Performance

Interpretation

Control-Flow Footprint Update

Current State of Business Process

Sub-Footprint Configs.

Control-FlowPerspective

BPMS 2 BPMS n

Trace State Interpretation

Open Traces

• Incrementally updates the Dynamic Footprint

• Only events are input (no enhancement of existing model) –

Control-Flow FP is exception: sub-footprints may be requested

through feedback loop

• Scalable: Constant amount of time (with regards to events and

traces)

• Conceptual FP update: FPn = (1-p)*FPn-1 + p*xe

Page 13: Introducing a Framework for Scalable Dynamic Process Discovery€¦ · Introducing a Framework for Scalable Dynamic Process Discovery David Redlich, Wasif Gilani, Thomas Molka, Marc

13/17 Motivation – Scalable Dynamic Process Discovery - Framework - Conclusion

BPMS 1

Run-time Processing of Standardized Events

Run-Time Event

Processing

Event Hub

Global, Standardized Events

...Events

from BPMS 3Events

from BPMS 2Events

from BPMS 1

Footprint Interpretation

Dynamic Control-flow Interpration

Enterprise System

Trace State Monitoring

Performance Footprint Update

Resource Footprint Update

ResourcePerspective

Performance Perspective

Current State of Traces

Reasoning on Current State

Dynamic Footprint

Control-FlowFootprint

ResourceFootprint

PerformanceFootprint

Dynamic Resource FP

Interpretation

Dynamic Performance

Interpretation

Control-Flow Footprint Update

Current State of Business Process

Sub-Footprint Configs.

Control-FlowPerspective

BPMS 2 BPMS n

Trace State Interpretation

Open Traces

Business Process State:

• Human-oriented

representation of the

BP State

• Basis for reasoning

techniques, e.g. Simulation

Footprint Interpretation:

• Not critical: less rigid

computation cost constraints

(low polynomial run-time)

• Execution scheduled or on-

demand

• Deterministic Algorithms

• Control-flow: Constructs

Competition Miner (CCM)

Page 14: Introducing a Framework for Scalable Dynamic Process Discovery€¦ · Introducing a Framework for Scalable Dynamic Process Discovery David Redlich, Wasif Gilani, Thomas Molka, Marc

Event Monitoring

Footprint Interpretation

14/17 Motivation – Scalable Dynamic Process Discovery - Framework - Conclusion

Process/Activity Lifecycle Transition

Footprint Type

Perspective Type

Start Schedule Assign Complete End

Open

Traces Global

Relations

Local

Relations

Resource

Associations

Instance

State

Control-Flow

Perspective

Resource

Perspective

Performance

Perspective

Single Entity

Performance

Lifecycle-FP-BPState-Mapping of

Framework Implementation

Page 15: Introducing a Framework for Scalable Dynamic Process Discovery€¦ · Introducing a Framework for Scalable Dynamic Process Discovery David Redlich, Wasif Gilani, Thomas Molka, Marc

Demo

15/17 Motivation – Scalable Dynamic Process Discovery - Framework - Conclusion

# of traces

BP’ - BP’

BP1 BP2

BP’m

BP in system:

Observed BP’: BP’nm+1 n-1

td ttr

BP’ - BP’1 m-1

tw

Warm-up, Detection, and Transition Time

Page 16: Introducing a Framework for Scalable Dynamic Process Discovery€¦ · Introducing a Framework for Scalable Dynamic Process Discovery David Redlich, Wasif Gilani, Thomas Molka, Marc

Conclusion

16/17 Motivation – Scalable Dynamic Process Discovery - Framework - Conclusion

Defined Scalable Dynamic Process

Discovery + Requirements/Characteristics

State of a Business Process: Dynamism and

Reflectivity (for different BP perspectives)

Application in TIMBUS for change detection

TIMBUS: research project for digitally

preserving business processes

Driven by requirements of real-life

industrial use-case (eHealth domain)

Page 17: Introducing a Framework for Scalable Dynamic Process Discovery€¦ · Introducing a Framework for Scalable Dynamic Process Discovery David Redlich, Wasif Gilani, Thomas Molka, Marc

Future Work

17/17 Motivation – Scalable Dynamic Process Discovery - Framework - Conclusion

Garbage Collection

Approximation for incomplete Footprint

information

Erasure of transition states (e.g. through

FP reset)

BP state vs. BP evolution

Incorporation of Data Perspective

Improve Generalization in Event-Hub (e.g.

event abstraction level)

Page 18: Introducing a Framework for Scalable Dynamic Process Discovery€¦ · Introducing a Framework for Scalable Dynamic Process Discovery David Redlich, Wasif Gilani, Thomas Molka, Marc

Thank you!

Email: [email protected]