Fast message passing for high performance computation in workstation clusters

被引:0
|
作者
Ou, X.M. [1 ]
Shen, J. [1 ]
Zheng, W.M. [1 ]
机构
[1] Dept. of Comp. Sci. and Technol., Tsinghua Univ., Beijing 100084, China
来源
Ruan Jian Xue Bao/Journal of Software | 2001年 / 12卷 / 03期
关键词
Communication - Network protocols - Parallel processing systems;
D O I
暂无
中图分类号
学科分类号
摘要
The message passing system is crucial to the computational performance of workstation clusters. With the fast growth in performance of networking device, the unnecessary overhead in traditional TCP/IP protocol has become the bottleneck of communication speed. To fully exploit the communication potential of the fast hardware, this paper introduces FMP, a new communication protocol specifically designed for Myrinet. FMP, which consists of a network part, is a concise protocol. The physical layer is assumed to be error free and thus the protocol becomes extremely simple and highly efficient. Compared with TCP/IP, the performance of FMP is much better. In this article the essential techniques used in the protocol are discussed and some results are presented.
引用
收藏
页码:317 / 322
相关论文
共 50 条
  • [1] Fast message passing communication for workstation clusters
    Shen, Jun
    Zheng, Wei-Min
    Chinese Journal of Advanced Software Research, 1999, 6 (03): : 236 - 246
  • [2] Low-latency message passing on workstation clusters using SCRAMNet
    Moorthy, V
    Jacunski, MG
    Pillai, M
    Ware, PP
    Panda, DK
    Page, TW
    Sadayappan, P
    Nagarajan, V
    Daniel, J
    IPPS/SPDP 1999: 13TH INTERNATIONAL PARALLEL PROCESSING SYMPOSIUM & 10TH SYMPOSIUM ON PARALLEL AND DISTRIBUTED PROCESSING, PROCEEDINGS, 1999, : 148 - 152
  • [3] Low-latency message passing on workstation clusters using SCRAMNet
    Moorthy, Vijay
    Jacunski, Matthew G.
    Pillai, Manoj
    Ware, Peter P.
    Panda, Dhabaleswar K.
    Page Jr., Thomas W.
    Sadayappan, P.
    Nagarajan, V.
    Daniel, Johns
    Proceedings of the International Parallel Processing Symposium, IPPS, 1999, : 148 - 152
  • [4] High-performance PVM based on fast message passing
    Xia, H.X.
    Zheng, W.M.
    Ruan Jian Xue Bao/Journal of Software, 2001, 12 (01): : 26 - 32
  • [5] SCALABLE HIGH PERFORMANCE COMPUTING ON FPGA CLUSTERS USING MESSAGE PASSING
    Creedon, Eoin
    Manzke, Michael
    2008 INTERNATIONAL CONFERENCE ON FIELD PROGRAMMABLE AND LOGIC APPLICATIONS, VOLS 1 AND 2, 2008, : 442 - 445
  • [6] Performance analysis of message-passing libraries on high-speed clusters
    Taboada, Guillermo L.
    Tourino, Juan
    Doallo, Ramon
    COMPUTER SYSTEMS SCIENCE AND ENGINEERING, 2010, 25 (01): : 63 - 78
  • [7] Performance analysis of Java']Java message-passing libraries on fast ethernet, myrinet and SCI clusters
    Taboada, GL
    Touriño, J
    Doallo, R
    IEEE INTERNATIONAL CONFERENCE ON CLUSTER COMPUTING, PROCEEDINGS, 2003, : 118 - 126
  • [8] Protocol-dependent message-passing performance on Linux clusters
    Turner, D
    Chen, XH
    2002 IEEE INTERNATIONAL CONFERENCE ON CLUSTER COMPUTING, PROCEEDINGS, 2002, : 187 - 194
  • [9] Message Passing for GPGPU Clusters: cudaMPI
    Lawlor, Orion Sky
    2009 IEEE INTERNATIONAL CONFERENCE ON CLUSTER COMPUTING AND WORKSHOPS, 2009, : 646 - 653
  • [10] High performance multithreaded message passing on a myrinet cluster
    Alves, A
    Pina, A
    Exposto, J
    Rufino, J
    APPLICATIONS OF HIGH-PERFORMANCE COMPUTING IN ENGINEERING VII, 2002, 7 : 241 - 250