| Summary: | 碩士 === 國立成功大學 === 資訊工程學系碩博士班 === 96 === In high performance routers design IP address lookup is a major bottleneck. Using multibit tries to represent IP routing tables for IP lookup is efficient together with pre-computation. The advantages of implementing IP lookup using multibit tries are simplicity and slight search time; and drawbacks of that are large memory usage and large update cost. To reduce memory usage, Controlled Prefix Expansion was proposed by Srinivasan and Varghese (1999), which was a pre-computation based on dynamic programming to determine the best strides of multibit tries, including fix-strides multibit tries and variable-stride multibit tries that makes the storage of these multibit tries are minimal. By such pre-compute algorithm we can simplify the data structure construction and thus archive fast lookup speed and improve memory requirement, but it didn’t concern about when a single prefix is added or deleted. Since instabilities in the backbone protocols can lead prefixes update, algorithms that can support fast updates are desirable.
In this thesis we proposed a pre-compute algorithm that can support fast updates by modifying Srinivasan’s algorithm. We explore optimization issues proposed in Srinivasan’s algorithm. Our proposed algorithm aims to reduce update overhead of multibit tries, and finds out the update optimal multibit tries solutions appropriate to a given update traffic. Compared the tries solution obtained by controlled prefix expansion (Srinivasan’s algorithm), our solutions reach a 30% reduction of update overhead when number of multibit trie level is eight.
|
|---|
