A Distributed Algorithm for GSPN Reachability Graph Generation

The applicability of generalized stochastic Petri nets (GSPNs) and other high-level modeling formalisms to real systems is often constrained by the explosion in the size of the underlying state-space representation. This paper describes a distributed program taking advantage of the large amount of memory and powerful computational resources of distributed computing systems to enable the construction of large state-space graphs generated by GSPN models. High state-space cardinalities and significant speedup with respect to sequential techniques are achieved by means of state space partitioning based on a distributed hashing function. The main algorithmic characteristics (global and local hashing and buffering of messages) are analyzed in detail and their effects are assessed by means of performance measurements on a PC cluster and a Cray T3D parallel machine. Performance results emphasize the good scalability with the number of processors of the proposed approach, even without fast interconnection networks. While the distributed reachability graph construction technique has been conceived for GSPN solution, the issues raised are relevant to other modeling systems whose common trait is the generation of state spaces with irregular and a priori unknown structure.

[1]  Shantanu Dutt,et al.  Scalable Global and Local Hashing Strategies for Duplicate Pruning in Parallel A* Graph Search , 1997, IEEE Trans. Parallel Distributed Syst..

[2]  Gianni Conte,et al.  Experiences on SIMD Massively Parallel GSPN Analysis , 1994, Computer Performance Evaluation.

[3]  Marco Ajmone Marsan,et al.  Generalized Stochastic Petri Nets: A Definition at the Net Level and Its Implications , 1993, IEEE Trans. Software Eng..

[4]  David M. Nicol,et al.  Distributed State Space Generation of Discrete-State Stochastic Models , 1995, INFORMS J. Comput..

[5]  Marco Ajmone Marsan,et al.  Modelling with Generalized Stochastic Petri Nets , 1995, PERV.

[6]  Gianni Conte,et al.  GSPN models of concurrent architectures with mesh topology , 1991, Proceedings of the Fourth International Workshop on Petri Nets and Performance Models PNPM91.

[7]  Graham Horton,et al.  State space construction and steady-state solution of GSPNs on a shared-memory multiprocessor , 1997, Proceedings of the Seventh International Workshop on Petri Nets and Performance Models.

[8]  Giovanni Rimassa,et al.  An Operating System Support to Low-Overhead Communications in NOW Clusters , 1997, CANPC.

[9]  Gianni Conte,et al.  Analysis and solution of generalized stochastic Petri nets: a data parallel CM-5 approach , 1995, HPCN Europe.

[10]  Giovanni Chiola,et al.  GreatSPN 1.7: Graphical Editor and Analyzer for Timed and Stochastic Petri Nets , 1995, Perform. Evaluation.

[11]  Vijay P. Kumar,et al.  Analyzing Scalability of Parallel Algorithms and Architectures , 1994, J. Parallel Distributed Comput..

[12]  Giovanni Chiola,et al.  Accelerating the Evaluation of Parallel Program Performance Models Using Distributed Simulation , 1994, Computer Performance Evaluation.

[13]  Marco Ajmone Marsan,et al.  GSPN Models of Markovian Multiserver Multiqueue Systems , 1990, Perform. Evaluation.

[14]  Gianni Conte,et al.  Analysis of large GSPN models: a distributed solution tool , 1997, Proceedings of the Seventh International Workshop on Petri Nets and Performance Models.

[15]  Marco Ajmone Marsan,et al.  A class of generalized stochastic Petri nets for the performance evaluation of multiprocessor systems , 1984, TOCS.

[16]  Dan C. Marinescu,et al.  A parallel algorithm for computing invariants of Petri net models , 1991, Proceedings of the Fourth International Workshop on Petri Nets and Performance Models PNPM91.

[17]  Sartaj Sahni,et al.  Performance metrics: keeping the focus on runtime , 1996, IEEE Parallel Distributed Technol. Syst. Appl..