DSVM based parallel Hash join algorithm and its performance evaluation

A new parallel Hash join algorithm based on Distributed Shared Virtual Memory(DSVM) called inter-subjoin parallel Hash is presented which fully exploits the advantages of DSVM and deals with the problem of dynamical imbalance very well. The benchmark is designed and implemented to evaluate and analyze the performance of the parallel join algorithms in a parallel and distributed object database system environment. The experiments have been done in the case of both even data distribution and Zipf skew data distribution. At the same time, it analyzes and compares the performance of the presented algorithm with other two algorithms. The experimental results show that for heavier workload a near linear speedup is gotten and parallel systems have a very good scalability built on DSVM in NOW environments.