Data Transfers between Processes in an SMP System: Performance Study and Application to MPI

Author

Buntinas, Darius ; Mercier, Guillaume ; Gropp, William

Author_Institution

Div. of Math. & Comput. Sci., Argonne Nat. Lab., IL

fYear

2006

fDate

14-18 Aug. 2006

Firstpage

487

Lastpage

496

Abstract

This paper focuses on the transfer of large data in SMP systems. Achieving good performance for intranode communication is critical for developing an efficient communication system, especially in the context of SMP clusters. We evaluate the performance of five transfer mechanisms: shared-memory buffers, message queues, the Ptrace system call, kernel module-based copy, and a high-speed network. We evaluate each mechanism based on latency, bandwidth, its impact on application cache usage, and its suitability to support MPI two-sided and one-sided messages

Keywords

cache storage; message passing; performance evaluation; queueing theory; shared memory systems; Ptrace system call; SMP system; cache usage; high-speed network; intranode communication; kernel module-based copy; large data transfers; message passing interface; message queues; performance study; shared-memory buffers; Application software; Bandwidth; Computer architecture; Computer science; Context; Delay; High-speed networks; Kernel; Laboratories; Mathematics;

fLanguage

English

Publisher

ieee

Conference_Titel

Parallel Processing, 2006. ICPP 2006. International Conference on

Conference_Location

Columbus, OH

ISSN

0190-3918

Print_ISBN

0-7695-2636-5

Type

conf

DOI

10.1109/ICPP.2006.31

Filename

1690653