Successfully reported this slideshow.

Rapid Performance Modeling by transforming Use Case Maps to Palladio Component Models

744 views

Published on

Published in: Technology, Education
  • Be the first to comment

  • Be the first to like this

Rapid Performance Modeling by transforming Use Case Maps to Palladio Component Models

  1. 1. Rapid Performance Modelingby transforming Use Case Mapsto Palladio Component ModelsChristian Vogel (KIT), Heiko Koziolek (ABB), Thomas Goldschmidt (ABB), Erik Burger (KIT)© ABB GroupSlide 1
  2. 2. MotivationPerformance Modeling at ABB Industrial Automation Domain Efficient control/data flowcrucial for real-time constraints Design sketches on whiteboards End-to-end latency as a requirement Challenges How to reason oncomplex control/data flows? How to break down global response timerequirements to performance budgetsfor software components?© ABB GroupSlide 2Sketchof a simplewebserver
  3. 3. Solution ApproachUCM2PCM Palladio Component Model (PCM) allowsconsideration and prediction ofperformance from the beginning But Control flow and structure in differentviews PCM models are difficult to createfor persons from industrial automation Thus Use Case Maps (UCM) as newfrontend for PCM Transformation from UCM to PCM© ABB GroupSlide 3
  4. 4. Related WorkModel transformations in SW Performance Engineering Performance engineering C. Smith and L.Williams, Addison-Wesley (2002)Performance Solutions: a practical guide to creating responsive, scalable software D. Menasce, L.W. Dowdy, V. Almeida, Prentice Hall (2004)Performance by design: computer capacity planning by example Survey of approaches for performance predictions S. Balsamo, A. Di Marco, P. Inverardi, and M. Simeoni, IEEE TSE (2004)Model-based performance prediction in software development: A survey H. Koziolek, Elsevier Performance Evaluation (2010)Performance evaluation of component-based software systems: A survey Transformation of UCM into Layered Queuing Networks (LQN) D. B. Petriu (2001)Layered software performance models constructed from use case map specifications Use of UML Marte Profile for creating performance models M. Woodside (2007)From annotated software designs (UML SPT/Marte) to model formalisms© ABB GroupSlide 4
  5. 5. FoundationsUse Case Maps Scenario-based modelinglanguage Part of the User RequirementsNotations (URN) Intuitive and easy to learn Graphical editor exists* Allows specification ofperformance annotations© ABB GroupSlide 5* http://lotos.site.uottawa.ca/ucm/bin/view/ProjetSEG/WebHome
  6. 6. UCM2PCMArchitecture© ABB GroupSlide 6
  7. 7. UCM2PCMPerformance Annotations added to UCMs© ABB GroupSlide 7Loop IterationCount Resource DemandsComponent AllocationBranch ProbabilitiesScenario Workload
  8. 8. UCM2PCMMapping Approach© ABB GroupSlide 8UCM PathComponentResponsibility
  9. 9. UCM2PCMMapping Approach© ABB GroupSlide 9Return CallComponent CallAll calls
  10. 10. UCM2PCMMapping Approach© ABB GroupSlide 10SystemRequiredCallSystemProvidedCallParentCall
  11. 11. UCM2PCMMapping Approach© ABB GroupSlide 11Usage Model
  12. 12. UCM2PCMLimitations Limited Input Assistance Performance Annotations need to be added as key-value pairs Limited PCM feature support Variables & Guarded Branches are not supported No Customized ResourceEnvironment supported Duplicated SEFFs For every call of a function a new SEFF is created© ABB GroupSlide 12
  13. 13. EvaluationGoals Evaluation of Accuracy Apply UCM2PCM to three test systems Validate the correctness of the PCM modelsresulting from the UCM2PCM transformation Measure and check the accuracy of the producedSimuCom performance results Evaluation of Usability Tutorial & Assignment for UCM2PCM Survey among developers for getting feedbackabout the UCM2PCM tool© ABB GroupSlide 13
  14. 14. EvaluationTested Models Media Store Plain Java web application for storing and retrievingmedia files. SPECjAppServer Industry-standard benchmark, designed to measure theperformance of application servers conforming to theJava EE 5.0 or later specifications. Business Reporting System Loosely modeled after a management informationsystem, formerly analyzed at Carlton University. Userscan retrieve live business data from the system and runstatistical analyses.© ABB GroupSlide 14
  15. 15. EvaluationBusiness Reporting System UCM© ABB GroupSlide 15
  16. 16. EvaluationAccuracy: Original vs. transformed UCM model Transformation successfully bridgedgap between UCM and PCM Simulation result difference toreference model below 15% andbelow 5% in most cases Accuracy is seen sufficiently forearly reasoning of different designalternatives© ABB GroupSlide 16
  17. 17. EvaluationUsability: qualitative user survey Main Advantages Comprehensibility ofmodel is high, also fornon-experts Fast(er) modeling Modeling complexsystems is seen criticalwith UCM2PCM Results are only hints,without statisticalrelevance© ABB GroupSlide 17
  18. 18. UCM2PCMConclusions and Future Work Transformation successfully bridgessemantic gap between UCM & PCM High result accuracy Mainly positive feedback by survey Future Work Adding Input Assistanceto UCM editor Reverse transformationfrom PCM to UCM Conducting furthercase studies & experiments© ABB GroupSlide 18
  19. 19. © ABB GroupApril 25, 2013 | Slide 19

×