Structure and Latency Analyses for High Performance Computing System Based on Asynchronous Optical Packet Switching

A novel high performance computing system based on optical packet switching and optical multicast technologies is presented. Distributed management architecture is used to alleviate the storing and computing pressures of every stage, which is easy to realize all-optical scalability. Asynchronous switching mode is accepted at every stage for high-speed and huge-capacity burst services transmission. The system scale and the stability features are analyzed, and a two stage system which interconnects 38,400 CPUs is adopted. Moreover, the average packets waiting latencies caused by the scheduling units and the recycling-fiber-delay-line based collision resolution units are simulated as 12.9ns and 0.63ns, respectively.