Greedy Prefix Cache for IP Routing Lookups

摘要

As the rapid growth of Internet and the communication link speed, it becomes increasingly challenging for network processors to timely route the incoming traffic to the destination ports. The traditional approach must look up the routing table based on the destination IP address to determine the output port. The ternary CAM approach provides fast associative look up, but is very costly for large routing tables. The trie-based algorithm allows inexpensive searching, but may not satisfy the growing speed requirement. Previous studies showed that the overall routing time can be shortened by adding a small prefix cache for the general trie-based routing algorithms. In caching the prefix, however, the nested prefixes are difficult to cache due to the constraint of the longest prefix matching requirement. This paper presents a greedy prefix caching technique to improve me prefix cache performance that allows caching the largest sub-tree of each prefix including the parent prefixes. Our experiment results show that the prefix cache using the proposed upgrade scheme can reduce the miss ratio by about 6-8% comparing to the best existing prefix caching mechanism.