Inter-Cluster Communication In NETRA
The focus of this chapter is inter-cluster communication in NETRA and performance evaluation of parallel algorithms when mapped across multiple clusters. When an algorithm is mapped on multiple clusters, processors belonging to different clusters may need to communicate, and therefore, inter-cluster communication is needed. However, unlike intra-cluster communication between processors, there may be conflicts in accessing the global interconnection network, global memory or other common resources. These conflicts need to be taken into account when computing the intercluster communication, and consequently, performance of algorithms when mapped across multiple clusters is affected by conflicts. In this chapter we present a method to evaluate inter-cluster communication time under the presence of conflicts. The method is based on the work by Patel [55,56].
KeywordsCluster Size Parallel Algorithm Interconnection Network Global Memory Communication Time
Unable to display preview. Download preview PDF.