A critical path tool for parallel simulation performance optimization

Describes a new technique for determining the critical path of a parallel simulation. Rather than relying on the traditional method of analyzing the dependencies in a directed graph, the new technique uses a metric called the earliest processing time (EPT) to determine the critical path. The new technique represents a slight innovation over the traditional method. This technique can be implemented either as a centralized algorithm or a distributed algorithm. As a centralized algorithm, the technique relies on a trace file containing the relevant EPTs the computation. As a distributed algorithm, this technique can be implemented so that the critical path becomes part of the output of the parallel computation. A centralized implementation of this technique has been developed and applied to a large and complex military-oriented simulation called CTLS (Concurrent Theater Level Simulation). Details of the two algorithms, and the results of applying the centralized algorithm to the CTLS simulation, are discussed in this paper.<<ETX>>

[1]  Susan L. Graham,et al.  Gprof: A call graph execution profiler , 1982, SIGPLAN '82.

[2]  Charles E. McDowell,et al.  Modeling Speedup greater than n , 1989, International Conference on Parallel Processing.

[3]  David R. Jefferson,et al.  Supercritical speedup , 1991, ANSS '91.

[4]  Barton P. Miller,et al.  Performance Measurement for Parallel and Distributed Programs: A Structured and Automatic Approach , 1989, IEEE Trans. Software Eng..

[5]  David R. Jefferson,et al.  Virtual time , 1985, ICPP.

[6]  David R. Jefferson,et al.  Supercritical speedup (discrete event simulation) , 1991, [1991] Proceedings of the 24th Annual Simulation Symposium.

[7]  Phil Hontalas,et al.  Distributed Simulation and the Time Wrap Operating System. , 1987, SOSP 1987.

[8]  Barton P. Miller,et al.  IPS-2: The Second Generation of a Parallel Program Measurement System , 1990, IEEE Trans. Parallel Distributed Syst..

[9]  K. Mani Chandy,et al.  Distributed computation on graphs: shortest path algorithms , 1982, CACM.

[10]  David R. Jefferson,et al.  The Performance of a Distributed Combat Simulation with the Time Warp Operating System , 1989, Concurr. Pract. Exp..

[11]  Horst F. Wedde,et al.  A critical path approach for testing distributed real-time systems , 1991, Proceedings of the Twenty-Fourth Annual Hawaii International Conference on System Sciences.