1 Challenges in Scaling E-Business Sites Menascé and Almeida. All Rights Reserved. Daniel A....

1

Challenges in Scaling E-Business Sites

Menascé and Almeida. All Rights Reserved.

Daniel A. MenascéDepartment of Computer Science

George Mason University

2

Impacts of Bad Performance

Bad performance: response time above 8 seconds (eight-second rule).

$43.5 billion lost each year in e-commerce due to bad performance (Zona Research, April 1999).

Holiday Season of 1998: over 1/3 of customers gave up due to slowness, 44% turned to conventional stores, 14% moved to another site.


3

Performance Problems for E-commerce tend to get worse!

Proliferation of mobile devicesEasier to use interfaces (VUI, wireless

and Web services on cars and airplanes, novel browsing paradigms)

Increasing load placed by agentsImpacts of authentication protocols

(e.g., TLS) on e-commerce site performance.


4

Typical QuestionsIs the online trading site prepared to

accommodate a 75% increase of trades/day?

Do I have enough servers to handle a peak demand 10x the average?

How fast can the site architecture be scaled up? What components should be upgraded? Database servers? Web servers? Application servers? Bandwidth?

How can I design a site that will meet its business goals?


5

Outline

ScalabilityA Reference Model for E-BusinessWorkload Characterization

Customer behavior model graphsClient/Server Interaction Diagrams

Concluding Remarks


6

A Reference Model for Electronic Business

BusinessModel

FunctionalModel

CustomerModel

ResourceModel

Characteristicsof the Business

NavigationalStructure of the

Site

Patterns ofCustomerBehavior

Site Architectureand ServiceDemands

External Metrics

Internal Metrics

Business View

TechnologicalView

Descriptors


7

External Metrics and Descriptors

Cover the Nature of BusinessMetrics:

Revenue throughput (dollars/sec)Potential lost revenue/secClick-to-look ratioLook-to-basket ratioBasket-to-buy ratioClick-to-buy ratioAvailabilityDownload timesPage views/dayUnique visitors/day


8

External Metrics and Descriptors

Descriptors:Number of registered customersNumber of potential customersMaximum number of simultaneous

customers in the storeNumber of items in the catalogEstimated operational cost


9

Workload Characterization

1. Determine the e-business functions made available by the site. Associate URLs or URL patterns to each e-business function.

2. Analyze the site’s HTTP logs to determine customer sessions.

3. Cluster customer sessions into groups of “similar” customer sessions, i.e., sessions that exhibit similar behavior.


10

home

browse

search

add select

pay

0.5

0.50.3

0.3

0.35

0.35 0.15

0.15 0.2

0.2

0.3

0.3

0.3

0.1

0.2

0.10.4

0.10.1

0.1 1.0

entry

1.0

Customer Behavior Model Graph


11

Metrics Derived from the CBMG

Average Number of Visits Per StateE.g., average number of searches per visit

to the site,

Average Buy (or open account) to Visit Ratio – also called conversion ratio.

Average Session Length Per Visit


12

Metrics Derived from the CBMG

1

1,

1n

kjkkj

entry

pVV

V


13

Workload Characterization Methodology

Merge and Filter

Get Sessions

Get CBMGs (clustering algorithm)

HTTP Logs

Session Log

CBMGs

Request Log


14

Result of Clustering Analysis

Cluster 1 2 3 4 5 6% of Sessions 44.28 28.00 10.60 9.29 6.20 1.50BV Ratio (%) 5.70 4.50 3.70 4.00 3.50 2.00Session Length 5.6 15 27 28 50 81AV Ratio (%) 11 15 21 20 32 50Vb+Vs 3.6 11.4 20 23 39 70

• Cluster 1: majority of sessions, short sessions, and highest BV ratio.

• Cluster 6: small fraction of sessions, large sessions, smallest BV ratio.


15

Buy to Visit Ratio vs. Session Length

y = 0.0003x2 - 0.07x + 5.7919

R2 = 0.931

0

1

2

3

4

5

6

7

0 10 20 30 40 50 60 70 80 90

Session Length


16

Architecture of E-Commerce Sites

Internet

E-business site

web andauthentication

servers

transactionservers

client

Mainframe:•legacy applications•enterprise data

LAN

LAN

Internet

E-business site

web andauthentication

servers

transactionservers

client

Mainframe:•legacy applications•enterprise data

LAN

LAN


17

RemarksFor e-commerce, need to characterize

the workload at higher levels of abstraction: sessions vs. requests.

Workload characterization has to be customer behavior-oriented as opposed to request-oriented.

Customer Behavior Model Graphs capture customer behavior and can be mapped to resource demands.


18

Capacity Planning and Performance Management:key to EC: the competitors are just a click

away!require predictive models: avoid ROTs!throwing more plumbing is not the

solution.models have to integrate the business,

customer, and resource aspects of the problem.

Remarks (cont’d)


19

Performance Models ofE-Commerce Sites

Daniel A. MenascéDept. of Computer ScienceGeorge Mason University

20

Component-level modelsserver

incominglink

outgoinglink

server

21


cpu

disk 1

disk 2

incominglink

outgoinglink

server

22


cpu

disk 1

disk 2

incominglink

outgoinglink

server

23

Component-level Models

Each component is represented by a resource (e.g. CPU, disk, communication link) and a queue of requests waiting for the resource.

resource

queue

24

Basic ConceptsUtilization of a resource:

Fraction of time the resource is busy serving requests during a measurement interval.

Example: the CPU was busy during 40 minutes during a measurement period of one hour. It’s utilization is then: 40/60 = 0.667 = 66.7%

Utilizations are measured as dimensionless numbers

25

Basic Concepts (cont’d)

A request may be using a resource (e.g., CPU, disk, etc) or waiting to use it.

The time spent using the resource does not depend on the number of resources waiting to use the resource.

The time spent waiting to use the resource depends on the load, i.e., on the number of requests in the queue waiting to use the resource.

26

Performance Model Parameters

Workload IntensityHTPP Requests/secTransactions/secE-business functions/sec

Service demands for each resource and each type of request.

27

Service Demandscpu

disk 1

disk 2

incominglink

outgoinglink

server

0.109 sec0.00107 sec

0.003 sec

0.08 sec

0.12 sec

Service demands do not include any queuing time! It is justservice time.

28

Computing Waiting Timescpu

disk 1

disk 2

incominglink

outgoinglink

server

0.109 sec0.00107 sec

0.003 sec

0.08 sec

0.12 sec

Waiting times depend on the load (arrival rate of requests) and on the service demands.

29

Di

Ri

i

ii D

DR

1'

Service demandat resource i

Utilization of resource i (Ui)

Computing Residence Times

Residence time at resource i

30

Residence Time at Incoming Linkcpu

disk 1

disk 2

incominglink

outgoinglink

Webserver

0.109 sec0.00107 sec0.003 sec

0.08 sec

0.12 sec req/sec

sec 00108.0

00107.051

00107.0

1'

IncLink

IncLinkIncLink D

DR

31

Residence Time at Outgoing Linkcpu

disk 1

disk 2

incominglink

outgoinglink

Webserver

0.109 sec0.00107 sec0.003 sec

0.08 sec

0.12 sec req/sec

sec 239.0

109.051

109.0

1'

OutLink

OutLinkOutlink D

DR

32

Residence Time at the CPUcpu

disk 1

disk 2

incominglink

outgoinglink

Webserver

0.109 sec0.00107 sec0.003 sec

0.08 sec

0.12 sec req/sec

sec 00305.0

003.051

003.0

1'

cpu

cpuCPU D

DR

33

Residence Time at Disk 1cpu

disk 1

disk 2

incominglink

outgoinglink

Webserver

0.109 sec0.00107 sec0.003 sec

0.08 sec

0.12 sec req/sec

sec 133.0

08.051

08.0

1 1

1'1

disk

diskdisk D

DR

34

Residence Time at Disk 2cpu

disk 1

disk 2

incominglink

outgoinglink

Webserver

0.109 sec0.00107 sec0.003 sec

0.08 sec

0.12 sec req/sec

sec 3.0

12.051

12.0

1 2

2'2

disk

diskdisk D

DR

35

Summary of Results

Average Response TimeSum of service demands

36

Response vs. Arrival Rate

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

1 2 3 4 5 6 7 8

Arrival Rate (requests/sec)

Res

po

nse

Tim

e (s

ec)

37

Example: online trading site

Open QN for the online trading site:

i

ii

lioShowPortfo

U

DR

RRRRRRR

1'

'6

'5

'4

'3

'2

'1

1 Challenges in Scaling E-Business Sites Menascé and Almeida. All Rights Reserved. Daniel A....

Documents

Transcript of 1 Challenges in Scaling E-Business Sites Menascé and Almeida. All Rights Reserved. Daniel A....