Title: Implications of application usage characteristics for collective communication offload

Authors: Ron Brightwell, Sue P. Goudy, Arun Rodrigues, Keith D. Underwood

Addresses: Sandia National Laboratories, P.O. Box 5800, MS-1110 Albuquerque, NM 87185-1110, USA. ' Sandia National Laboratories, P.O. Box 5800, MS-1110 Albuquerque, NM 87185-1110, USA. ' Sandia National Laboratories, P.O. Box 5800, MS-1110 Albuquerque, NM 87185-1110, USA. ' Sandia National Laboratories, P.O. Box 5800, MS-1110 Albuquerque, NM 87185-1110, USA

Abstract: The global, synchronous nature of some collective operations implies that they will become the bottleneck when scaling to hundreds of thousands of nodes. One approach improves collective performance using a programmable network interface to directly implement collectives. While these implementations improve micro-benchmark performance, accelerating applications will require deeper understanding of application behaviour. We describe several characteristics of applications that impact collective communication performance. We analyse network resource usage data to guide the design of collective offload engines and their associated programming interfaces. In particular, we provide an analysis of the potential benefit of non-blocking collective communication operations for MPI.

Keywords: MPI; non-blocking; collective communication; resource usage; resource management; network interface; offload engines.

DOI: 10.1504/IJHPCN.2006.010633

International Journal of High Performance Computing and Networking, 2006 Vol.4 No.3/4, pp.104 - 116

Published online: 10 Aug 2006 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article