Adaptive Linear Prediction of Queues for Reduced Rate Scheduling in Optical Routers

Abstract—This paper describes a switching scheme that can be used to reduce delays in optical routers. Optical switches have a larger bandwidth and can potentially use less power than electronic switches. However op-tical switches take a significant amount of time (of the order of sec) to change from one configuration to another. This introduces a time lag be-tween the scheduling decision and actual change in configuration, lead-ing to higher delays. We propose a scheduling scheme that decides the schedule for the future using maximum weight matching (MWM) on the predicted lengths of the virtual output queues (VOQs). The prediction is done using the adaptive Least Mean Square (LMS) algorithm. We show through simulations for correlated arrivals that this scheme reduces delay significantly in comparison to a scheme that makes a scheduling decision for the future using MWM on the current lengths of virtual output queues. I