Efficient Support for Multicomputing on ATM Networks

摘要: AbstractThe emergence of a new generation 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 remote access dictates granularityof parallelism that can be supported. Thus, in addition traditional distribution mechanisms such asRPC, clusters should support lightweight communication paradigms for executing parallelapplications.This paper describes simple model notion memory access.Applications one host perform direct read or write operations user-definedremote buffers. We have implemented prototype system using com-mercially available workstations and ATM networks. Our uses kernel-based emulation ofremote instructions, through unused processor opcodes; thus, applications(or runtime libraries) see machine access. show modelcan supportedsafely efficiently current systems; forexample, 40-byteremote writeoperationcompletes 30