Performance Analysis of a Hardware Solution for Collective Communication Operations on Clusters of PCs
Resumo
This paper presents a performance evaluation of a hardware solution for some collective communication operations on a cluster of 8 PCs based on Myrinet and FastEthernet switches. It was implemented a modified version of TTL_PAPERS [DIE 95] for 8 nodes and the hardware collective operations were included in MPICH version 1.1 [GRO 96], which is an implementation of MPI message passing standart. The performance evaluation was done measuring the execution time of the collective operations and evaluating their impact on the total execution time of several applications programs. We concluded the special hardware for collective communication operations is a suitable option to improve the performance on a parallel machine based on cluster of Pcs.
Palavras-chave:
Collective communication operations, MPI, performance evaluation, cluster of PCs
Referências
BALAY, S. et al. PETSc 2.0 User Manual. Mathematics and Computer Science Division, Argonne National Laboratory, 1995. [link].
DIETZ, H. G. et al. Purdue’s Adapter for Parallel Execution and Rapid Synchronization: the TTL_PAPERS Design. West Lafayette, School of Electrical Engineering Purdue University, Jan. 1995. (Technical Report). [link]
DONGARRA, J. J.; WALKER, D. W. Software Libraries for Linear Algebra Computations on High Performance Computers. SIAM Review, v. 37, n. 2, p. 151–180, June 1995.
GROPP, W. et al. A high-performance implementation of the MPI message passing interface standard. Parallel Computing, v. 22, p. 789–828, 1996.
NUPAIROJ, N.; NI, L. M. Benchmarking of multicast communication services. East Lansing, Department of Computer Science Michigan State University, Apr. 1995. (Technical Report MSU-CPS-ACS-103).
SNIR, M. et al. MPI: the complete reference. London: The MIT Press, 1996.
DIETZ, H. G. et al. Purdue’s Adapter for Parallel Execution and Rapid Synchronization: the TTL_PAPERS Design. West Lafayette, School of Electrical Engineering Purdue University, Jan. 1995. (Technical Report). [link]
DONGARRA, J. J.; WALKER, D. W. Software Libraries for Linear Algebra Computations on High Performance Computers. SIAM Review, v. 37, n. 2, p. 151–180, June 1995.
GROPP, W. et al. A high-performance implementation of the MPI message passing interface standard. Parallel Computing, v. 22, p. 789–828, 1996.
NUPAIROJ, N.; NI, L. M. Benchmarking of multicast communication services. East Lansing, Department of Computer Science Michigan State University, Apr. 1995. (Technical Report MSU-CPS-ACS-103).
SNIR, M. et al. MPI: the complete reference. London: The MIT Press, 1996.
Publicado
24/10/2000
Como Citar
TORRES, Martha; KOFUJI, Sergio Takeo.
Performance Analysis of a Hardware Solution for Collective Communication Operations on Clusters of PCs. In: INTERNATIONAL SYMPOSIUM ON COMPUTER ARCHITECTURE AND HIGH PERFORMANCE COMPUTING (SBAC-PAD), 12. , 2000, São Pedro/SP.
Anais [...].
Porto Alegre: Sociedade Brasileira de Computação,
2000
.
p. 115-120.
DOI: https://doi.org/10.5753/sbac-pad.2000.41211.
