In this paper, an index-based parallel scheduler for optical burst switching (OBS) networks, which can find feasible voids on different channels in parallel with O(1) time complexity, achieve the highest possible efficiency, and is suitable for efficient hardware implementation, is proposed. The channel scheduler proposed consists of two phases: searching the feasible voids on all data channels and selecting optimal void among feasible voids found. In the first phase, an index vector of voids for each channel is constructed by dividing the scheduling time window of the channel into $N$ slots, and an index-based searching approach is applied in parallel to each channel to find the feasible voids by a few simple index based bit operations and once memory reading and comparing operations. A configurable optimal void selection approach, which selects optimal void according to the external configurable signal by using a configurable comparand translator and a general selection process, is designed to supported different scheduling algorithms such as LAUC-VF, Min-SV, Max-EV, and Best-Fit with full efficiency. Formulae for important design parameters are given. The hardware architecture of the index based parallel scheduler is designed, and a 16-channel scheduler with the full resource utilization of LAUC-VF is implemented in FPGA (Field Programmable Gate Array) with a scheduling time of 33.3 ns per burst request. The support to different scheduling algorithms is also validated experimentally.
© 2011 IEEEPDF Article