Previous [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12]

Journal of Information Science and Engineering, Vol.18 No.5, pp.815-824 (September 2002)


Hash Join Algorithms on SMPs Clusters: Effects of
Netcaches on its Scalability and Performance

Edward David Moreno Ordonez
Department of Computer Science
Euripides Foundation of Marilia
CEP 17525-901, Marilia, S.P., Brazil
E-mail: edmoreno@fundanet.br

We investigate the effect that caches, particularly caches for remote accesses, have on the performance of hash join algorithms. The join is a computationally intensive operation of relational databases and is used in many important applications. Thus, there are a considerable number of studies on the parallel hash join. However, most of the previous research does not show how cache affects the performance of these algorithms. In this paper, we show the impact and benefits of remote caches (Netcaches) on the overall performance of parallel hash join algorithms running on SMP clusters. Furthermore, we show the effects of these caches on speedup and scalability of these algorithms. Our simulation results leads us to conclude that the execution time of hash join algorithms on moderns multiprocessors, with large local and remote caches, could be reduced up to 70%. Finally, we show results for verifying the big effects of netcaches on scalability of these algorithms.

Keywords: hash join algorithms, netcaches, SMPs, clusters, performance evaluation

Full Text () Retrieve PDF document (200209_10.pdf)

Received September 22, 2001; accepted April 15, 2002.
Communicated by Jang-Ping Sheu, Makato Takizawa and Myongsoon Park.