PB_Cube: a join algorithm on Hypercube Systems

This paper presents and analyzes a join algorithm for cube-connected multiprocessors. The algorithm, called PB_Cube, was developed to exploit as much as possible the hypercube structural characteristics, enabling us to achieve synchronization between processors, maximum parallelism and a reduction of the traffic on the network. Moreover, it doesn't present any bucket overflow and buffer overflow problems. Experimental results indicate the algorithm has linear speedup and scale-up in a significant range. Furthermore its performance is comparable with that of Hybrid Hash [9] join and better than that of Parallel Distributive algorithm [4]. Our algorithm has the advantage of being able to handle non-equijoin. These results become even more interesting when we consider that the costs for Hybrid Hash and Parallel Distributive algorithms are given under hypothesis stronger than ours.