Efficient Support for Multicomputing on ATM Networks

AbstractThe emergence of a new generation of networks will dramatically increase the attractiveness ofloosely-coupled multicomputers based on workstation clusters. The key to achieving high performancein this environment is efficient network access, because thecost of remote access dictates the granularityof parallelism that can be supported. Thus, in addition to traditional distribution mechanisms such asRPC, workstation clusters should support lightweight communication paradigms for executing parallelapplications.This paper describes a simple communication model based on the notion of remote memory access.Applications executing on one host can perform direct memory read or write operations on user-definedremote memory buffers. We have implemented a prototype system based on this model using com-mercially available workstations and ATM networks. Our prototype uses kernel-based emulation ofremote read and write instructions, implemented through unused processor opcodes; thus, applications(or runtime libraries) see direct machine support for remote memory access. We show that this modelcan be supportedsafely and efficiently on current systems; forexample, a 40-byteremote writeoperationcompletes in 30