Performance evaluation of offloading software modules to cluster network

Authors:
Keiichi Aoki;Hiroki Maruoka;Koichi Wada;Masaaki Ono
Affiliations:
Department of computer science, University of Tsukuba, Tsukuba, Ibaraki, Japan;Department of computer science, University of Tsukuba, Tsukuba, Ibaraki, Japan;Department of computer science, University of Tsukuba, Tsukuba, Ibaraki, Japan;Department of computer science, University of Tsukuba, Tsukuba, Ibaraki, Japan
Venue:
PDCN'07 Proceedings of the 25th conference on Proceedings of the 25th IASTED International Multi-Conference: parallel and distributed computing and networks
Year:
2007

Citing 8
Cited 0

MPI-SIM: using parallel simulation to evaluate MPI programs

Proceedings of the 30th conference on Winter simulation
SPINE: a safe programmable and integrated network environment

Proceedings of the 8th ACM SIGOPS European workshop on Support for composing distributed applications
PM2: a high performance communication middleware for heterogeneous network environments

Proceedings of the 2000 ACM/IEEE conference on Supercomputing
EMP: zero-copy OS-bypass NIC-driven gigabit ethernet message passing

Proceedings of the 2001 ACM/IEEE conference on Supercomputing
Myrinet: A Gigabit-per-Second Local Area Network

IEEE Micro
MPI and Embedded TCP/IP Gigabit Ethernet Cluster Computing

LCN '02 Proceedings of the 27th Annual IEEE Conference on Local Computer Networks
NIC-based offload of dynamic user-defined modules for Myrinet clusters

CLUSTER '04 Proceedings of the 2004 IEEE International Conference on Cluster Computing
TCP offload is a dumb idea whose time has come

HOTOS'03 Proceedings of the 9th conference on Hot Topics in Operating Systems - Volume 9

Quantified Score

Hi-index	0.00

Visualization

Abstract

A design of software to offload user-defined software modules to Maestro2 cluster network, named Maestro dynamic offloading mechanism(MDO), and results of its performance evaluation are described. Maestro2 is a high-performance network for clusters. The network interface and the switch of Maestro2 have a general-purpose processor tightly coupled with dedicated communication hard-ware. MDO enables the users to offload software modules to both the network interface and the switch. The effectiveness of MDO are discussed through the evaluation of offloading collective communications and parallel Gauss-Jordan elimination benchmark.