TY - GEN
T1 - Static round-robin dispatching schemes for Clos-network switches
AU - Pun, Konghong
AU - Hamdi, M.
PY - 2002
Y1 - 2002
N2 - The Clos-network is widely recognized as a scalable architecture for high-performance switches and routers. One of the key challenges in designing a Clos-network switch for a high-speed environment is the design of the dispatching/scheduling so as to be efficient for a wide range of traffic patterns, yet practical to be implemented in hardware. Based on the static round-robin scheduling technique, we propose the SRRD cell dispatching algorithm and its variants for Clos-network switches in this paper. Our algorithms are based on the request-grant-accept (RGA) handshaking scheme, which can be implemented using simple distributed arbiters at the input and output of the Clos-network. The intuition behind our SRRD schemes is to desynchronize the pointers of the arbiters in a static way and by to use a rotating-search technique to improve the performance under non-uniform traffic. Our simulation results demonstrate that our algorithms achieve the lowest delay and highest throughput among all other related schemes. In addition, their hardware implementations seem to be even simpler than that of related algorithms.
AB - The Clos-network is widely recognized as a scalable architecture for high-performance switches and routers. One of the key challenges in designing a Clos-network switch for a high-speed environment is the design of the dispatching/scheduling so as to be efficient for a wide range of traffic patterns, yet practical to be implemented in hardware. Based on the static round-robin scheduling technique, we propose the SRRD cell dispatching algorithm and its variants for Clos-network switches in this paper. Our algorithms are based on the request-grant-accept (RGA) handshaking scheme, which can be implemented using simple distributed arbiters at the input and output of the Clos-network. The intuition behind our SRRD schemes is to desynchronize the pointers of the arbiters in a static way and by to use a rotating-search technique to improve the performance under non-uniform traffic. Our simulation results demonstrate that our algorithms achieve the lowest delay and highest throughput among all other related schemes. In addition, their hardware implementations seem to be even simpler than that of related algorithms.
UR - http://www.scopus.com/inward/record.url?scp=84905494962&partnerID=8YFLogxK
U2 - 10.1109/HPSR.2002.1024258
DO - 10.1109/HPSR.2002.1024258
M3 - Conference contribution
AN - SCOPUS:84905494962
SN - 488552184X
SN - 9784885521843
T3 - IEEE International Conference on High Performance Switching and Routing, HPSR
SP - 329
EP - 333
BT - HPSR 2002 - Workshop on High Performance Switching and Routing
PB - IEEE Computer Society
T2 - 2002 Workshop on High Performance Switching and Routing: Merging Optical and IP Technologies, HPSR 2002
Y2 - 26 May 2002 through 29 May 2002
ER -