PUMA

PERFORMANCE FROM UNIFIED MODEL ANALYSIS

Web page: www.sce.carleton.ca/rads/puma/

The PUMA project beginning in Nov. 2002 will develop a unified approach to building performance models from design models that specify scenarios for Use Cases.

Progress Report June 29 2004: Eight presentations in a Zip file.

Highlights of the Plan

• Scenario input based on the OMG UML Profile for Schedulability, Performance and Time, accepted in June 2002.
• Other forms of definition of scenarios, either based on UML, or on other languages such as Use Case Maps.
• A Scenario Definition Interface which is adaptable to other specification tools
• A Core Scenario Model in a language to be defined, to integrate a wide variety of scenario specifications.
• Automated generation of a meta-model for performance
• A Core Performance Model in a language to be defined, which is adaptable to generation of the widest possible variety of performance model formats.
• Performance modeling using existing tools.
• Initially, Layered Queueing, and Queueing Network model formats for ensured scalability.
• Simulation.
• Design evaluation by model experimentation, and feedback of results as measures attached to the design as
• design suggestions,
• identified hot spots,
• software resource analysis.

Invitation

The potential value of the Core Model (Core Scenario Model, Core Performance Model, and transformation between them) is to create a standard interface between different software specification tools and different performance models. We would like to work together with others to create a widely acceptable and useful definition of the Core Model. If you are interested, please get in touch with us. We would like to form a group who will participate and contribute. One possible goal is some kind of standard for tool interworking.

Metamodel of the Core Scenario Model

The MOF metamodel as UML diagrams, and the XML schema for the Core Scenario Model (CSM) are available here.

Resources and Background Reading

We have transformed UML design models into layered queueing performance models:

• Gordon Gu and D. C. Petriu, "XSLT Transformation from UML Models to LQN Performance Models", Proc. of 3rd Int. Workshop on Software and Performance WOSP'2002, pp.227-234, Rome, Italy, July 2002.

• D.C.Petriu, H.Shen, "Applying the UML Performance Profile: Graph Grammar based derivation of LQN models from UML specifications", (pdf) in Computer Performance Evaluation - Modelling Techniques and Tools, (Tony Fields, Peter Harrison, Jeremy Bradley, Uli Harder, Eds.) Lecture Notes in Computer Science 2324, pp.159-177, Springer Verlag, 2002.

Many of our ideas about scenario modeling are based on our experience with Use Case Maps. A lot of background on UCMs, and the UCM Navigator tool, will be found HERE. This includes a proposal called URN for an ITU standard. The Navigator can be downloaded and is available uner an open-source license.

Previously we have generated performance models from UCMs with the UCM2LQN tool, we have applied it to study early design issues, and we have studied performance of different concurrency options:

• Dorin Petriu, Murray Woodside, "Software Performance Models from System Scenarios in Use Case Maps", Proc. 12 Int. Conf. on Modelling Tools and Techniques for Computer and Communication System Performance Evaluation (Performance TOOLS 2002), London, April 2002. PDF
• Dorin Petriu, Murray Woodside, "Analysing Software Requirements Specifications for Performance", Proc. Third Int. Workshop on Software and Performance, Rome, July 2002.  PDF
• Khalid Siddiqui, Murray Woodside, "Performance -Aware Software Development (PASD) Using Resource Demand Budgets", Proc. Third Int. Worshop on Software and Performance, Rome, July 2002. PDF
• C. Scratchley, C. M. Woodside, "Evaluating Concurrency Options in Software Specifications", Proc 7th Int. Symp. on Modeling, Analysis and Simulation of Computer and Telecomm Systems (MASCOTS99), College Park, Md., pp 330 - 338, October 1999.  PostscriptPDF

Our performance modeling is done using the Layered Queueing formalism, which is well suited to modeling software resources. Information on layered queueing, including the Layered Queueing Network Solver (LQNS) can be be found HERE. LQNS can be obtained by contacting Woodside.

Investigators

• Murray Woodside (PI), Dorina Petriu, Carleton University {cmw|petriu}@sce.carleton.ca
• Daniel Amyot, University of Ottawa , damyot@site.uottawa.ca

Collaborating Companies: we have participation of individuals from

• Artisan Software, Mitel, Motorola, Nortel, Rational