期刊文献+

基于RapidIO的高性能通信接口的设计与实现 被引量:4

DESIGN AND IMPLEMENTATION OF A HIGH PERFORMANCE COMMUNICATION INTERFACE BASED ON RAPIDIO
在线阅读 下载PDF
导出
摘要 在高性能计算领域中,集群系统对于高速互连网络的要求越来越高。通过研究将RapidIO用于集群系统的高速互连网络,设计并且实现了基于RapidIO的高性能通信接口,该接口针对MPI高度优化,并且充分利用RapidIO的硬件优势,具有基于优先级的多流、可靠有序的数据包传输等特点,并且使用了空闲缓存池提高性能。实验数据表明,在带宽和延迟上,该专用接口都比原来的接口有优势。 In the field of high performance computation, there is growing concerns with the effect of high speed network performance on cluster system. In this article, it analyses the application of RapidlO in high speed network of cluster system, designs and implements the high performance communication Interface based on RapidlO. This special Interface aims at highly optimizing MPI, makes the best use of RapidlO' s hardware advantage, has the characters of multi-stream communication with priority scheduling strategy, reliable and orderly data transferring ability. Moreover, it uses buffer pool to improve the efficiency. From the results of the tests, it can tell that this special Interface shows considerable advantages comparing with its original Interface in both high bandwidth and low latency.
出处 《计算机应用与软件》 CSCD 2009年第7期43-45,共3页 Computer Applications and Software
基金 国家高技术研究发展计划(2005AA104031)
关键词 RAPIDIO MPI 集群 高性能计算 消息传递接口 RapidlO MPI Cluster High performance computing Message passing interface
作者简介 梁基,硕士,主研领域:高性能计算,Linux集群。
  • 相关文献

参考文献8

  • 1TOP500 Supercomputing Sites[EB/OL].http://www.top500.org/.
  • 2Pulleyblank W.Terascale computing and BlueGene.Proceedings of the 21st International Conference on Computer Design[C].2003,10.
  • 3RapidIO Trade Association.RapidIO Interconnect Specification[M].Rev 2002,1(2).
  • 4Message Passing Interface Forum[EB/OL].http://www.mpi-forum.org/.
  • 5Argonne National Laboratory.http://www-unix.mcs.anl.gov/mpi/mpich.
  • 6Huang W,Santhanaraman G,Jin H W,et al.Design and implementation of high performance MVAPICH2:MPI2 over InfiniBand,Int′l Sympsoium on Cluster Computing and the Grid (CCGrid)[J].Singapore,May 2006.
  • 7RapidIO Trade Association.RapidIO Interconnect Specification Part II:Message Passing Logical Specification[M].Rev 2002 1.2,(06).
  • 8Kamal H,Penoff B,Wagner A.SCTP versus TCP for MPI.In Supercomputing '05:Proceedings of the 2005 ACM/IEEE conference on Supercomputing[J],Washington,D.C.,USA,2005.

同被引文献29

引证文献4

二级引证文献7

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