Optimizing Communications of Data Parallel Programs in Scalable Cluster Systems

摘要

With the improvement of personal computers and high-speed network, clusters have become the trend in designing high performance computing environments. As the researches of relative hardware and software technology of Cluster and Grid are constantly improved, the application of Cluster is growing popular. Due to information progress and increasing calculation capacity required by all kind of applications; the calculation has also extended to cross-network calculation. Through the Internet, which connects several Clusters, the mass calculation platform is combined into Cluster Grid. As a result of data partition and exchange that may happen during the executing program, communication localization becomes important in programming efficiency. This paper, then, proposes a mathematical method which achieves excellent data partitioning and maintains data calculation in local environment. We also conduct some theoretical analysis on the amounts of computing nodes and data partition in the hope of being applied to practical parallel environment and further to reduce communication cost.