Traffic scheduling in non-blocking optical packet switches with minimum delay

For performance guaranteed OPS switches with reconfiguration overhead, it has been shown that packet delay can be minimized by using N switch configurations (where N is the switch size) to schedule the traffic. However, this usually involves an exorbitant speedup requirement, which makes it impractical under current technology. In this paper, a new minimum-delay scheduling algorithm QLEF (quasi largest-entry-first) is proposed. We prove that QLEF pushes the required speedup bound to the lowest known level. As an example, when N=950, QLEF only requires a speedup of S_schedule=21.33 instead of 42.25 for MIN (B. Towles and W.J. Dally, 2003) and 30.27 for aⁱ-SCALE (B. Wu and K.L. Yeung, 2005). This gives a 50% improvement over MIN and 30% over aⁱ-SCALE