Maximizing MPI Point-to-Point Communication Performance on RDMA-enabled Clusters with Customized Protocols

被引:7
作者
Small, Matthew [1 ]
Yuan, Xin [1 ]
机构
[1] Florida State Univ, Dept Comp Sci, Tallahassee, FL 32303 USA
来源
ICS'09: PROCEEDINGS OF THE 2009 ACM SIGARCH INTERNATIONAL CONFERENCE ON SUPERCOMPUTING | 2009年
关键词
MPI; RDMA; Rendezvous Protocol;
D O I
10.1145/1542275.1542320
中图分类号
TP3 [计算技术、计算机技术];
学科分类号
0812 ;
摘要
Message Passing Interface (MPI) point-to-point communications are usually realized with two protocols, the eager protocol for small messages and the rendezvous protocol for medium and large sized messages. Traditional sender-initiated rendezvous protocols are sub-optimal in many situations. In this work, we propose to refine the rendezvous protocol for medium and large messages on RDMA-enabled clusters with three protocols that are customized for different situations, a hybrid protocol for medium sized messages when the sender arrives early, a sender-initiated protocol for large messages when the sender arrives early, and a receiver-initiated protocol when the receiver arrives early. In comparison to traditional sender-initiated rendezvous protocols, the proposed scheme reduces unnecessary synchronizations, decreases the number of control messages that are in the critical path of communications, and improves the communication progress, which results in a significantly better communication-computation overlap capability. We present and analyze these protocols, and describe how these protocols and the eager protocol can be seamlessly integrated in one system without introducing an excessive number of control messages. We have implemented the proposed scheme for InfiniBand clusters. The experimental results demonstrate the effectiveness of the proposed technique.
引用
收藏
页码:306 / 315
页数:10
相关论文
共 17 条
[1]  
Amerson G, 2004, INT C CLUST COMP
[2]  
[Anonymous], MPI MESS PASS INT ST
[3]  
[Anonymous], NAS Parallel Benchmarks
[4]  
[Anonymous], 2003, INFINIBAND HOST CHAN
[5]  
[Anonymous], MVAPICH: MPI over InfiniBand
[6]  
DANALIS A, 2008, GRAVEL COMMUNICATION
[7]  
KE J, 2005, LNCS, V2648, P19
[8]  
KEPPITIYAGAMA C, 2000, INT C COMM COMP
[9]  
KEPPITIYAGAMA C, 2001, IEEE INT PAR DISTR P
[10]  
KUMAR R, 2008, LOCK FREE ASYNCHRONO