Using multiple hash functions to improve IP lookups

Abstract
High performance Internet routers require a mechanism for very efficient IP address look-ups. Some techniques used to t his end, such as binary search on levels, need to construct quickly a good hash table for the appropriate IP prefixes. In this paper we describe an appr oach for obtaining good hash tables based on using multiple hashes of each input key (which is an IP address). The methods we describe are fast, simple, scalable, parallelizable, and flexible. In particular, in i nstances where the goal is to have one hash bucket fit into a cache line, using mult iple hashes proves extremely suitable. We provide a general analysis of this hashing technique and specifically discuss its application to binar y search on levels. Keywords—IP routing, hashing, binary search on levels, two choices.

This publication has 10 references indexed in Scilit: