Search

In reference one, a network architecture called hyper-cluster employing wavelength-routing, multi-hop packet switching and optical reconfiguration is presented. It is modularly scalable to very large configurations on both hardware and operational bases. A hyper-cluster uses a logical hierarchy for the purpose of addressing but guarantees that all access nodes have a constant number of transceivers. It is a cluster of regular graphs; the clustering structure follows the traffic distribution in a grand granularity. The issue of operational scalability is addressed by presenting a scalable routing protocol and a scalable reconfiguration protocol. When using shuffle-net as the building block, a novel routing scheme called quantified deflection routing is presented. The scheme improves call blocking performance significantly. In this paper, we propose a distributed reconfiguration protocol. The network throughput and virtual call blocking performance is obtained via simulation on large networks (with size beyond 200 nodes). Numerical results show that the dynamic self- routing protocol, combined with quantified deflection routing for shuffle-net, can achieve excellent resource utilization efficiency for very large networks. When the call arrival rate is below 0.3, the capacity provided by the hyper-cluster dynamic routing algorithm is close to that of an infinite capacity centralized switch (lowest possible call blocking caused exclusively by congestion on the finite capacity user input/output links, never by the switch fabric itself).© (1997) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.