Fast routing lookups are crucial for the forwarding performance of IP routers.Longest prefix match makes routing lookups difficult.This paper proposes a parallel routing lookup system that applies a method to partition a routing table.The method can divide all prefixes in a routing table into several prefix sets where prefixes don't overlap.Based on the method

this paper also presents a common parallel lookup framework that reduces "longest prefix matching" in all the prefixes to "only prefix matching" in several prefix sets.The framework can effectively simplify the design of lookup algorithms and improve lookup performance.The framework is suitable for most lookup algorithms.For simple binary search algorithm

our system can reach log

2

(2N/B) lookup complexity (where N is prefix number in a rou

ting table and B is an integer bigger than 4).Also

the system can be scaled to IPv6 easily.