Hybrid Analytical-Simulation Model Used to Evaluate and Improve System Performance

Performance prediction has been intensively studied in the last decade, alongside the accelerated development of distributed systems. This paper focuses on a hybrid approach regarding model solving, combining two popular prediction techniques applied separately so far, analytical and simulation modeling, in order to benefit from the strengths of both. The input UML model with MARTE (Modeling and Analysis of Real-time and Embedded systems) annotations is transformed into a hierarchically decomposed performance model, and performance results for simulated sub models are used by an analytical solver. The validation of the proposed method is to be performed with a tool called PHYMSS (Performance Hybrid Model Solver and Simulator) developed by the authors that implements both the hybrid solver and a multithreaded simulator.

[1]  Raffaela Mirandola,et al.  Deriving a queueing network based performance model from UML diagrams , 2000, WOSP '00.

[2]  C. Murray Woodside,et al.  Performance analysis of distributed server systems , 2000 .

[3]  Lothar Thiele,et al.  Performance evaluation of network processor architectures: combining simulation with analytical estimation , 2003, Comput. Networks.

[4]  Vittorio Cortellessa,et al.  From UML models to software performance results: an SPE process based on XML interchange formats , 2005, WOSP '05.

[5]  Mary K. Vernon,et al.  Approximate Mean Value Analysis for Closed Queuing Networks with Multiple-Server Stations , 2007 .

[6]  Hessam S. Sarjoughian,et al.  Hybrid discrete event simulation with model predictive control for semiconductor supply-chain manufacturing , 2005, Proceedings of the Winter Simulation Conference, 2005..

[7]  Simonetta Balsamo,et al.  UML-PSI: the UML performance simulator , 2004, First International Conference on the Quantitative Evaluation of Systems, 2004. QEST 2004. Proceedings..

[8]  J. George Shanthikumar,et al.  A Unifying View of Hybrid Simulation/Analytic Models and Modeling , 1983, Oper. Res..

[9]  Ioan Jurca,et al.  Towards early performance assessment based on UML MARTE models for distributed systems , 2009, 2009 5th International Symposium on Applied Computational Intelligence and Informatics.

[10]  Murray Woodside.  Tutorial introduction to layered modeling of software performance , 2002 .

[11]  K. Mani Chandy,et al.  Linearizer: a heuristic algorithm for queueing network models of computing systems , 1982, CACM.

[12]  Jan van der Wal,et al.  Simulation and or (operations research) in combination for practical optimization , 2005, Proceedings of the Winter Simulation Conference, 2005..

[13]  C. Murray Woodside,et al.  A Metamodel for Generating Performance Models from UML Designs , 2004, UML.

[14]  Robert G. Sargent.  A historical view of hybrid simulation/analytic models , 1994, Proceedings of Winter Simulation Conference.

[15]  Filip De Turck,et al.  Hybrid performance modeling approach for network intensive distributed software , 2007, WOSP '07.

[16]  Vittorio Cortellessa,et al.  A Framework for Automated Generation of Architectural Feedback from Software Performance Analysis , 2007, EPEW.

[17]  Connie U. Smith,et al.  A performance model interchange format , 1999, J. Syst. Softw..

[18]  Stephen S. Lavenberg,et al.  Mean-Value Analysis of Closed Multichain Queuing Networks , 1980, JACM.

[19]  Ioan Jurca,et al.  Phymss: performance hybrid model solver and simulator based on UML MARTE diagrams , 2010, WOSP/SIPEW '10.