Efficient layering for high speed communication: the MPI over Fast Messages (FM) experience
作者： Mario Lauria , Scott Pakin , Andrew Chien
关键词: Layering 、 Computer science 、 Myrinet 、 Scheduling (computing) 、 Embedded system 、 Flow control (data)
摘要: We describe our experience of designing, implementing, and evaluating two generations high performance communication libraries, Fast Messages (FM) for Myrinet. In FM 1, we designed a simple interface provided guarantees reliable in-order delivery, flow control. While this was significant improvement over previous systems, it not enough. Layering MPI atop 1 showed that only about 35% the bandwidth could be delivered to higher level APIs. Our second generation layer, 2, addresses identified problems, providing gather-scatter, interlayer scheduling, receiver control, as well some convenient API features which simplify programming. 2 can deliver 55–95% APIs such MPI. This is especially impressive absolute bandwidths have increased fourfold 90 MB/s. general issues encountered in matching layers, solutions embodied 2.
springer.com
sci-hub.se 参考文章(28)
J. Postel, Transmission Control Protocol Internet Request for Comment (RFC793). ,vol. 793, pp. 1- 91 ,(1981)
Cezary Dubnicki, Angelos Bilas, Yuqun Chen, Stefanos Damianakis, Kai Li, VMMC-2 : Efficient Support for Reliable, Connection-Oriented Communication Hot Interconnect'97. ,(1997)
Matt Welsh, Thorsten von Eicken, Anindya Basu, Incorporating Memory Management into User-Level Network Interfaces Cornell University. ,(1997)
J. Postel, User Datagram Protocol RFC768. ,vol. 768, pp. 1- 3 ,(1980)
L.A. Giannini, A.A. Chien, A software architecture for global address space communication on clusters: put/get on fast messages high performance distributed computing. pp. 330- 337 ,(1998) , 10.1109/HPDC.1998.710019
Bert J. Dempsey, Alfred C. Weaver, W. Timothy Strayer, Xtp: The Xpress Transfer Protocol ,(1992)
Scott Pakin, Mario Lauria, Andrew Chien, High Performance Messaging on Workstations: Illinois Fast Messages (FM) for Myrinet conference on high performance computing (supercomputing). pp. 55- ,(1995) , 10.1145/224170.224360
Vijay Karamcheti, John Plevyak, Andrew A. Chien, Runtime Mechanisms for Efficient Dynamic Multithreading Journal of Parallel and Distributed Computing. ,vol. 37, pp. 21- 40 ,(1996) , 10.1006/JPDC.1996.0105
Mengjou Lin, David H. C. Du, Jenwei Hsieh, Joseph P. Thomas, James A. MacDonald, Distributed network computing over local ATM networks conference on high performance computing (supercomputing). pp. 154- 163 ,(1994) , 10.5555/602770.602802
Vijay Karamcheti, Andrew A. Chien, A comparison of architectural support for messaging in the TMC CM-5 and the Cray T3D international symposium on computer architecture. ,vol. 23, pp. 298- 307 ,(1995) , 10.1145/223982.224440