Performance forecasting: towards a methodology for characterizing large computational applications

We present a methodology that can identify and formulate performance characteristics of a computational application and uncover program performance trends on very large, future computer architectures and problem sizes. Based on this methodology we present "performance forecast diagrams" that predict the scalability of a large seismology application suite on a terabyte data set. We find that the applications scale well up to a large number of processors, given an interconnection network similar to the one of the SGI/Cray Origin architecture. However we find that if we increase the computation-to-communication speed ratio by a factor of 100, the different applications of the seismic suite start exhibiting architectural "sweet spots", at which the communication overhead starts to dominate computation time. The presented methodology has proven to be useful in characterizing large computational applications. It is being applied in a project to create a repository of realistic programs and their characteristics.

[1]  R. Eigenmann,et al.  Hierarchical processors-and-memory architecture for high performance computing , 1996, Proceedings of 6th Symposium on the Frontiers of Massively Parallel Computation (Frontiers '96).

[2]  Barton P. Miller,et al.  The Paradyn Parallel Performance Measurement Tool , 1995, Computer.

[3]  B. Miller,et al.  The Paradyn Parallel Performance Measurement Tools , 1995 .

[4]  Alan Jay Smith,et al.  Analysis of benchmark characteristics and benchmark performance prediction , 1996, TOCS.

[5]  Gheith A. Abandah,et al.  Configuration independent analysis for characterizing shared-memory applications , 1998, Proceedings of the First Merged International Parallel Processing Symposium and Symposium on Parallel and Distributed Processing.

[6]  Insung Park,et al.  URSA MAJOR: Exploring Web Technology for Design and Evaluation of High-Performance Systems , 1998, HPCN Europe.

[7]  Rudolf Eigenmann,et al.  Benchmarking with real industrial applications: the SPEC High-Performance Group , 1996 .

[8]  Daniel A. Reed,et al.  Experimental Analysis of Parallel Systems: Techniques and Open Problems , 1994, Computer Performance Evaluation.

[9]  Ken Kennedy,et al.  A static performance estimator to guide data partitioning decisions , 1991, PPOPP '91.

[10]  Gheith A. Abandah,et al.  Modeling the communication performance of the IBM SP2 , 1996, Proceedings of International Conference on Parallel Processing.

[11]  Quinn Snell,et al.  HINT: A new way to measure computer performance , 1995, Proceedings of the Twenty-Eighth Annual Hawaii International Conference on System Sciences.

[12]  Alan Jay Smith,et al.  Performance Characterization of Optimizing Compilers , 1992, IEEE Trans. Software Eng..

[13]  Thomas J. LeBlanc,et al.  Parallel performance prediction using lost cycles analysis , 1994, Proceedings of Supercomputing '94.

[14]  VALERIE TAYLOR.  A Simulation-based Coste ciency Study of Hierarchical Heterogeneous Machines for Compiler-and Hand-Parallelized Applications , 1997 .

[15]  Mark J. Clement,et al.  Performance prediction of PVM programs , 1996, Proceedings of International Conference on Parallel Processing.