{"title":"DCA: a distributed CCA framework based on MPI","authors":"Felipe Bertrand, R. Bramley","doi":"10.1109/HIPS.2004.1299193","DOIUrl":null,"url":null,"abstract":"The CCA is a component architecture for high-performance scientific applications. In this architecture components are parallel entities that are connected directly or in a distributed manner. The problem of communication between scientific parallel programs with differing numbers of processes is called the \"M/spl times/N problem\". This paper discusses problems and solutions regarding the M/spl times/N problem in the context of the CCA. We also present a prototype implementation of a distributed CCA framework with M/spl times/N capabilities. This implementation reuses many MPI concepts and constructions to build the parallel-remote port invocation mechanism. Leveraging MPI helps developers that are familiar with that communication library and benefits from its performance and high degree of scalability.","PeriodicalId":448869,"journal":{"name":"Ninth International Workshop on High-Level Parallel Programming Models and Supportive Environments, 2004. Proceedings.","volume":"1 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2004-04-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"38","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Ninth International Workshop on High-Level Parallel Programming Models and Supportive Environments, 2004. Proceedings.","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/HIPS.2004.1299193","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 38
Abstract
The CCA is a component architecture for high-performance scientific applications. In this architecture components are parallel entities that are connected directly or in a distributed manner. The problem of communication between scientific parallel programs with differing numbers of processes is called the "M/spl times/N problem". This paper discusses problems and solutions regarding the M/spl times/N problem in the context of the CCA. We also present a prototype implementation of a distributed CCA framework with M/spl times/N capabilities. This implementation reuses many MPI concepts and constructions to build the parallel-remote port invocation mechanism. Leveraging MPI helps developers that are familiar with that communication library and benefits from its performance and high degree of scalability.