Decision Tree Based Algorithm for IPv6 Routing Table Lookup

Abstract

The next generation Internet Protocol, IPv6, has attracted growing attention and has been deploying worldwide, especially in Europe, Asia, and North America. With 128-bit address, IPv6 provides an extremely large address space and poses a great challenge to the routing table lookup algorithms. The previous algorithms need reconsideration. An IPv6-specific longest prefix matching (LPM) scheme called DTLU (Decision Tree LookUp) is introduced is this paper. By dividing route prefixes into chunks, the scheme converts the IPv6 LPM problem to the multi-dimensional packet classification problem and then uses the decision tree to conquer the classification problem. Moreover observations are made on the real live IPv6 backbone BGP routing tables and on the converted classifiers. Then the observations are used to enhance the basic DTLU. DTLU supports incremental updates and only needs tens of kilobytes memory for the current real live IPv6 backbone BGP routing tables. The evaluation results show that the sample software implementation of DTLU can achieve more than 10 million lookups per second on a PC with Pentium4 2.4GHz CPU, 512M DDR333 memory, and Linux operating system.