TY - GEN
T1 - Support for multiple classes of traffic in multicomputer routers
AU - Rexford, Jennifer
AU - Shin, Kang G.
N1 - Publisher Copyright:
© 1994, Springer Verlag. All Rights Reserved.
PY - 1994
Y1 - 1994
N2 - Emerging parallel real-time and multimedia applications broaden the range of performance requirements imposed on the interconnection network. This communication typically consists of a mixture of different traffic classes, where guaranteed packets require bounds on latency or throughput while good average performance suffices for the best-effort traffic. This paper investigates how multicomputer routers can capitalize on low-latency routing and switching techniques for best-effort traffic while still supporting guaranteed communication. Through simulation experiments, we show that certain architectural features are best-suited to particular performance requirements. Based on these results, the paper proposes and evaluates a router architecture that tailors low-level routing, switching, and flow-control policies to the unique needs of best-effort and guaranteed traffic. Careful selection of these policies, coupled with fine-grain arbitration between the classes, allows the guaranteed and best-effort packets to share network bandwidth without sacrificing the performance of either class.
AB - Emerging parallel real-time and multimedia applications broaden the range of performance requirements imposed on the interconnection network. This communication typically consists of a mixture of different traffic classes, where guaranteed packets require bounds on latency or throughput while good average performance suffices for the best-effort traffic. This paper investigates how multicomputer routers can capitalize on low-latency routing and switching techniques for best-effort traffic while still supporting guaranteed communication. Through simulation experiments, we show that certain architectural features are best-suited to particular performance requirements. Based on these results, the paper proposes and evaluates a router architecture that tailors low-level routing, switching, and flow-control policies to the unique needs of best-effort and guaranteed traffic. Careful selection of these policies, coupled with fine-grain arbitration between the classes, allows the guaranteed and best-effort packets to share network bandwidth without sacrificing the performance of either class.
UR - http://www.scopus.com/inward/record.url?scp=84893777657&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84893777657&partnerID=8YFLogxK
U2 - 10.1007/3-540-58429-3_32
DO - 10.1007/3-540-58429-3_32
M3 - Conference contribution
AN - SCOPUS:84893777657
SN - 9783540584292
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 116
EP - 130
BT - Parallel Computer Routing and Commumcation - 1st International Workshop, PCRCW 1994, Proceedings
A2 - Bolding, Kevin
A2 - Snyder, Lawrence
PB - Springer Verlag
T2 - 1st International Workshop on Parallel Computer Routing and Communication, PCRCW 1994
Y2 - 16 May 1994 through 18 May 1994
ER -