Introduction to Wimax

1
A Model-driven Approach to Describe and Predict
the Performance of Composite Services
Authors Andrea DAmbrogio and Paolo
Bocciarelli Department of Computer Science,
Systems and Production, University of Roma Tor
Vergata, Roma, Italy. Published
2007 Presentation for Course (SYSC 510) Professor
Majumdar
Presented By Pragash Rajeswaran
2
Presentation Outline

• Introduction
• Related Works
• Focus of the research document
• Concept Overview
• Model-driven Performance Prediction of Composite
  Services
• Example Application
• Implementation Validation Issues
• Conclusion
• QA

3
Introduction

• Distributed applications are rapidly converging
  towards the adoption of a computing paradigm
  based on service-oriented architectures (SOA).
• Service Is a self-describing, open component
  that supports rapid composition of distributed
  applications.
• Composite service Can be seen as a set of
  services that cooperate to execute a process that
  defines the collaboration workflow.
• Service providers problems
• Problems of predicting the level of quality
  of service (QoS) that can be offered
• to service consumers
• Dynamically managing the QoS at execution
  time
• Why service providers face the above problems?
• Interested to describe the performance
  characteristics of offered services
• To predict the level of performance that can be
  offered to service consumers
• May want to explore several alternatives
  configurations of composite service during
  design and execution process

4
Related Works

• Several approaches have been proposed in
  literature that dealt with performance aspects of
  web services and their composition. Some of the
  related work are given below
• Queuing-based models have been used to identify
  performance problems in web services or web
  applications.
• LQN models are used to calculate the response
  time of a clinical decision support system based
  on web services.
• Performance models of LQN type are used to
  mitigate performance penalties that have to be
  paid when exposing a legacy application as a web
  service.
• LQN models are used to evaluate the impact of web
  services on the overall system performance.
• Contributions that deals at a larger extent
  (i.e., not limited to queuing-based modeling)
• An approach to trigger re-planning opportunities
  during composite service execution is proposed.
• Simulation techniques are used for analyzing the
  performance of composite web services.

5
Focus of research document

• Introduced a model-driven approach for
  integrating performance prediction into service
  composition processes carried out by use of
  Business Process Execution Language (BPEL) for
  Web Services.
• In addition it focuses on,
• Use of LQN models for predicting the performance
  of composite web services
• Introduces a method for automatically building a
  LQN model from UML models of BPEL-based service
  composition processes.
• Characteristic of the method introduced
• Method makes use of a WSDL extension that allows
  service providers to describe the performance
  characteristics of the services they offer.
• Method can be fully automated
• Method could be applied both to static and
  dynamic composition of web services.
• Static composition The services to be composed
  are chosen at design time,
• Dynamic composition The services are chosen at
  execution time. ( LQN models can be used as an
  on-line tool system)

6
Presentation Outline

• Introduction
• Related Works
• Focus of the research document
• Concept Overview
• Composite Service
• BPEL
• P-WSDL
• Model-driven Performance Prediction of Composite
  Services
• Example Application
• Implementation Validation Issues
• Conclusion
• QA

7
Composite Services Overview

• Composite service can be seen as a set of
  services that cooperate to execute a process that
  defines the interaction workflow.
• Two different styles of collaboration between
  services
• Orchestration represents the control from the
  perspective of one party, which is the central
  authority (or coordinator) that controls the
  execution of component web services.
• Choreography much more collaborative and does
  not rely on a central coordinator that controls
  the execution of the overall process.
• Overview of Business Process Execution language
  (BPEL)
• BPEL represents a convergence of the ideas in
  XLANG (a block structured process language) and
  WSFL (a graph-based process language)
  specifications, which have been superseded by the
  BPEL specification.
• BPEL process is an XML document typically
  generated with graphical design tools by business
  analysts rather than programmers. (BPEL process
  is executed by an orchestration)
• BPEL is intended for modeling two types of
  processes abstract and executable processes.
• BPEL is essentially a layer on top of WSDL, with
  WSDL defining the specific operations allowed and
  BPEL defining how the operations can be
  sequenced.

8
P-WSDL Overview

• WSDL description is an XML document that contains
  all the information about service capabilities
  and invocation mechanisms.
• WSDL document only addresses the functional
  aspects of a web service without containing any
  useful description of non-functional or quality
  of service characteristics.
• WSDL extension has been introduced in for the
  description of performance characteristics of a
  web service, such as response time, throughput,
  utilization, etc

The proposed approach is lightweight, so that
existing WSDL documents can easily be extended
without altering their original content, and
automated, in order to effectively support its
adoption.
9
P-WSDL Overview Cont


10
Presentation Outline

• Introduction
• Related Works
• Focus of the research document
• Concept Overview
• Model-driven Performance Prediction of Composite
  Services
• Example Application
• Implementation Validation Issues
• Conclusion
• QA

11
Model-driven Performance Prediction of Composite
Services

• Steps
• 1 - Composite service is initially specified as a
  workflow of abstract services. The composite
  service specification produces as output an
  abstract model.
• 2 - Service discovery is then carried out to bind
  each abstract service to one of the several
  concrete services matching the abstract service
  interface.
• Two alternative service selection approaches
• Local optimization approach
• Global planning approach.
• 3 - The annotated AD is then transformed both
  into a performance model of LQN type (LQN) and
  into a BPEL document.
• 4 - The LQN model is given as input to the LQN
  solver that carries out the model evaluation step
  and yields as output the predictions about the
  performance of the composite web service.

12
Model-driven Performance Prediction of Composite
Services Cont
13
Convert BPEL to LQN
14
Presentation Outline

• Introduction
• Related Works
• Focus of the research document
• Concept Overview
• Model-driven Performance Prediction of Composite
  Services
• Example Application Travel Planning Composite
  Service
• Implementation Validation Issues
• Conclusion
• QA

15
Travel Planning Service

• Travel Planning System Specification
• Looking for flight and hotel room availability
  (according consumer request)
• Obtaining information about car renting.
• The travel plan should also include information
  about transportation from the airport to the
  hotel. (According to user preferences)
• The plan may include either a timetable of
  airport shuttles or an estimated cab fare.
• The travel plan is finally presented to the
  customer for approval and booking.
• Note
• Assumption an average 70 of customers prefer a
  cab rather than an airport shuttle.
• The performance requirement on the overall
  composite web service is specified in terms of
  the response time of the entire scenario, which
  should not exceed 3 seconds.

16
Travel Planning example Cont

• The abstract model of the composite web service
  are build.
• Service discovery activity is then carried out
  to bind each abstract service to a specific
  concrete service matching the abstract service
  interface.
• Performance data annotated by use of the SPT
  profile are extracted from the P-WSDL documents
  of component web services.
• Table 2 summarizes the sizes of input/output
  messages, for each invocation of external web
  services.

The abstract model is then transformed to an
executable model, as illustrated by the annotated
AD in this figure 8 (Given at next Slide).
17
Travel Planning example Cont
18
Travel Planning LQN Model
19
Impact of Two TM Services
Question The service provider may want to
explore the impact on overall performance of two
alternative transportation manager (TM) services
that provide different performance
characteristics.
This table summarizes performance data extracted
from the PWSDL documents of the two alternative
services, namely TMA and TMB.
20
Impact of Two TM Services Cont

• This figure 9 illustrates,
• The performance prediction obtained by
• evaluating the LQN models of the two alternative
  composite web services.
• It is easy to see that the composite service that
  binds the flight manager service to TMB is more
  scalable than the one using TMA.

• Note
• The composite web service with TMA satisfies the
  established performance requirements for a number
  of users not exceeding 25.
• In this case, the transportation manager service
  itself is the system bottleneck, as shown in
  Table 4,
• Which describes the utilization of invoked
  services in the two alternatives for 25 users of
  the composite web services.

21
Presentation Outline

• Introduction
• Related Works
• Focus of the research document
• Concept Overview
• Model-driven Performance Prediction of Composite
  Services
• Example Application
• Implementation Validation Issues
• Conclusion
• QA

22
Implementation Validation Issues

• The use of a standard data format is
  essential for enabling model interoperability, or
  the easy interchange of models between
  distributed and heterogeneous tools and
  repositories.
• The required transformations can be
  specified effectively by use of declarative or
  imperative model transformation languages, and
  then executed by use of a transformation engine
  that takes as input the source XML file with the
  transformation rules and yields as
• output the target XML file.
• The main advantage of this solution
• The transformation rules are specified at an
  abstract level
• Automatically mapped to a specific
  implementation technology
• Without specifying details about how the
  transformation has to be carried out
• Without implementing a transformation engine
  from scratch.
• Note
• Full-featured transformation engines are not yet
  available
• The transformations in the proposed approach
  have been implemented by use of standard
  XML-based languages and tools, such as XSLT.

23
Implementation Validation Issues Cont..

• Prototype Validating process
• The component web services that implement the
  core functions of various
• BPEL processes have been developed by use of the
  Java platform and are
• Locally hosted, as well as the resulting
  composite web service.
• Existing load on component web services has been
  emulated by use of multi-threaded application
  clients acting as existing service consumers.
• The approach has been validated by comparing the
  response time and the
• utilization level predicted by the model with
  real data monitored on the composite
• web service.
• Testing Implementation not included on the
  paper
• Static service composition in real world
  environment
• Dynamic service composition was not tested
• Nevertheless, the high degree of
  automation and interoperability of the proposed
  approach can be considered as a first step toward
  such an attractive solution

24
Conclusion

• Key points
• A model-driven approach for integrating
  performance prediction into service composition
  processes described by use of BPEL is introduced.
• The proposed approach is founded on P-WSDL
  (Performance-enabled WSDL)
• A performance-oriented extension of WSDL, the
  language for describing
• the information about service capabilities and
  invocation mechanisms.
• The main feature of the approach is the use of a
  model transformation method to
• automatically build LQN models of BPEL processes
  described by use of appropriately
• annotated UML models.
• The method enables service providers to
  effectively predict if the composite service
• meets the performance requirements formalized in
  the service level agreement.
• The approach currently support static service
  composition and is illustrated by use
• of an example application to a composite web
  service for travel planning.
• Future work
• An extended validation of the proposed approach,
  by use of publicly hosted web services
• Its extension to other QoS attributes, by use of
  QoS-aware descriptions of component
• web services (Dynamic Service composition)

25
Presentation for Course SYSC 5101 Professor
Majumdar
Pragash Rajeswaran