Pseudo-Banyan Optical WDM Packet Switching System With Near-Optimal Packet Scheduling

Abstract

We present a novel pseudo-Banyan optical packet switching system (SBOPSS) for optical wavelength division multiplexing (WDM) networks. The system includes a group of pseudo-Banyan space switches together with single-stage downsized fiber-delay-line-based optical buffers. SBOPSS is scalable, with the result that each pseudo-Banyan space switch performs packet switching only for a cluster of wavelengths. The downsized optical buffers that are shared by output ports via the use of a small number of internal wavelengths result in efficient reduction in packet loss. Essentially, SBOPSS employs a packet scheduling algorithm, referred to as the parallel and incremental packet scheduler (PIPS). Given a set of newly arriving packets per time slot, PIPS determines a maximum number of valid paths (packets) to be scheduled with the current buffers' state taken into account. The algorithm aims at maximizing the system throughput subject to satisfying three constraints, which are switch-contention free, buffer-contention free, and sequential delivery. Significantly, we prove that PIPS is incremental in the sense that the computed-path sets are monotonically non-decreasing over time. We then propose a hardware parallel system architecture for the implementation of PIPS. As is shown, PIPS achieves a near-optimal solution with an exceptionally low computational complexity, O, (|P| X log(2), NMW)), where P is the newly-arriving-packet set, N the number of input ports, and M and W the numbers of internal and external wavelengths, respectively. From simulation results that pit the PIPS algorithm against four other algorithms, we show that PIPS outperforms these algorithms on both system throughput and computational complexity.