Arbitration-free Time-Division Permutation Switching suitable for All-Optical Implementation

摘要

Pursuing high-bandwidth and transparency for optical communications, we propose here a multistage switching fabric whose internal interconnection patterns are periodically reconfigured. Specifically, by cascading guided-wave modules containing each a small set of interleaved, independent addressable plane-to-plane global interconnections, we experimentally demonstrate arbitration-free time-division permutation routing in a transparent multistage architecture suitable for high-bandwidth inter-processor communications in massively parallel machines. Furthermore -and perhaps more interestingly- computer simulations confirms that the addition of small inter-stage buffering nodes (suitable for being implemented as optical fiber delay lines) along with a simple and local packet flow arbitration rule (adapted to the global, periodic interconnection reconfiguration strategy) would lead to fairly good network performance for packet switching applications.