Adjusted Weight-Balanced Binary Tree for IP Routing Table Lookup*

Abstract Various algorithms have recently been proposed for Internet routers to improve the performance of routing table lookup. However, most of previous research effort does not utilize the access frequencies of route entries. Since Internet traffic has significant temporal localities [9]-[11], one could largely improve the performance of routing table lookup by taking advantage of access frequencies. An example that aims at minimizing the average lookup time can be found in [12]. Since global optimization is computationally prohibited, the authors suggested to construct a binary tree based on a weight-balancing heuristic. We present in this paper that the resulted weight-balanced binary tree can be adjusted with a simple algorithm to further improve the performance. The adjustment algorithm is so simple that it basically does not increase the complexity in constructing the binary tree. Through analysis, we found that the improvement can be as large as 25% for a tree with four leaf nodes. For a tree of arbitrary size, the improvement could be larger than 12.5%.