Parallel routing table computation for scalable IP routers

Abstract

The exponential growth of Internet traffic requires that routers be scalable. A generic scalable IP router is typically composed of a number of routing nodes (RNs) interconnected by a scalable switching fabric. A critical issue in the design of scalable IP routers is to provide a global and consistent routing table in such distributed-memory architectures. A parallel routing table computation approach is proposed in this paper. By dividing an OSPF area into a number of disjoint Within-Area-Routing-Regions (WARRs), the computation required to calculate the routing table can be divided into multiple independent portions and done by different RNs in parallel. Compared to conventional approaches, this approach can have a speedup between n and n2 for within-area link state updates and n for outside-area link state updates, where n is the number of RNs in the router. This parallel routing table computation approach requires no modification of OSPF.