期刊文献+

一种MPI并行软件的动态演化框架

A FRAMEWORK OF DYNAMIC EVOLUTION FOR MPI PARALLEL SOFTWARE
下载PDF
导出
摘要 为使MPI并行计算软件具有运行时改变软件功能和拓扑的能力,基于MPI系统模型提出一种用于描述和实现动态演化的两层软件框架。在管理层,采用适用于消息传递的管道-过滤器风格规范系统的结构、分析并行软件的可演化性,并约定一种相应的动态ADL用于描述演化行为。在计算层,设计符合管道-过滤器风格的构件元模型及过滤器的重配置管理自动机。针对实现演化的两个核心问题,即进程的启动和数据连接的建立释放,实现基于MPI动态进程管理的重配置方法。通过性能测试,验证了该软件框架的可用性,为并行计算软件的动态演化提供了一种新的实现方法。 To make MPI parallel computing software have the capabilities of changing software functions and topology, based on MPI system model we proposed a two-layer software framework for describing and implementing dynamic evolvable. On management layer, we normalised the system structure with pipe-filter style applicable to massage passing, and analysed the evolution property of parallel applications, as well as agreed on a correlated dynamic ADL for evolution behaviours description. On computation layer, we designed the meta-model of components suitable for pipe-filter style and the finite automaton (FA) for controlling reconfiguration of filter. In order to solving two key challenges of evolution implementation, i. e. , the launching of process and the establishing and freeing of data connection, the reconfigure method based on MPI dynamic process management was realised. Through performance test the feasibility of this framework was verified. This framework provides a new implementation method of dynamic evolution for parallel computing software.
出处 《计算机应用与软件》 CSCD 2016年第1期8-10,76,共4页 Computer Applications and Software
关键词 软件工程 动态演化 并行计算 MPI 体系结构描述语言 Software engineering Dynamic evolution Parallel computing Message passing interface (MPI) Architecture description language (ADL)
  • 相关文献

参考文献10

  • 1蔡志明,王希敏.软件声纳的概念与趋势[J].声学技术,2007,26(5):968-971. 被引量:10
  • 2Wang H,Potluri S,Bureddy D,et al.GPU-Aware MPI on RDMA-Enabled Clusters:Design,Implementation and Evaluation[J].IEEE Transactions on Parallel and Distributed Systems,2014,25(10):2595-2605.
  • 3王怀民,史佩昌,丁博,尹刚,史殿习.软件服务的在线演化[J].计算机学报,2011,34(2):318-328. 被引量:25
  • 4Rajesh S,Calvin J R.Re SHAPE:A Framework for Dynamic Resizing and Scheduling of Homogeneous Applications in a Parallel Environment[C]//China:2007 International Conference on Parallel Processing,2007:44-64.
  • 5George B,Thomas H,Ala R.On Scalability for MPI Runtime Systems[C]//USA:2011 IEEE International Conference on Cluster Computing,2011:187-195.
  • 6Naughton T,Engelmann C,Vallee G,et al.Supporting the Development of Resilient Message Passing Applications Using Simulation[C]//Italy:22nd Euromicro International Conference on Parallel,Distributed and Network-Based Processing,2014:271-278.
  • 7Castain R,Woodall T,Daniel D,et al.Recent Advances in Parallel Virtual Machine and Message Passing Interface[M].German:Springer,2005.
  • 8谢仲文,李彤,代飞,秦江龙,张璇,莫启,朱锐.基于Petri网的面向动态演化的软件体系结构建模[J].计算机应用与软件,2012,29(10):36-39. 被引量:7
  • 9Pavan B,Darius B,David G,et al.PMI:A Scalable Parallel ProcessManagement Interface for Extreme-Scale Systems[C]//German:In Proceedings of the 17th European MPI Users’Group Meeting Conference on Recent Advances in the Message Passing Interface,2010:31-41.
  • 10Argonne National Laboratory.PMI-2 API[EB/OL].USA:University of Illinois,(2012-11-10).[2013-3-16].http://wiki.mcs.anl.gov/mpich2/index.php/PMI_v2_API.

二级参考文献19

共引文献39

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部