Towards a Communication Framework based on Balanced Message Flow Distribution

Abstract

Applications running in grids and other large scale distributed systems have several communication requirements which are not easy to satisfy by the existing communication middleware, like high throughput data transfer, reliable point-to-point messaging and unreliable multicast streaming. In this paper we propose a scalable, fault tolerant communication framework based on the peer-to-peer paradigm. The framework's design is oriented towards high volume point-to-point and multicast data delivery. The proposed system uses balanced message flow distribution over multiple paths in order to achieve high throughput and the lexicographic distance between the peer's names in order to guarantee that the message flow converges to destination.