Reliable multicast fault tolerant MPI in the Grid environment

Abstract

Grid environments have recently been developed with low stretch and overheads that increase with the logarithm of the number of nodes in the system. Getting and sending data to/from a large numbers of nodes is gaining importance due to an increasing number of independent data providers and the heterogeneity of the network/Grid. One of the key challenges is to achieve a balance between low bandwidth consumption and good reliability. In this paper we present an implementation of a reliable multicast protocol over a fault tolerant MPI: MPICHV2. It can provide one way to solve the problem of transferring large chunks of data between applications running on a grid with limited network links. We first show that we can achieve similar performance as the MPICH-P4 implementation by using multicast with data compression in a cluster. Next, we provide a theoretical cluster organization and GRID network architecture to harness the performance provided by using multicast. Finally, we present the conclusion and future work.