Bit-Shuffled Trie: IP Lookup with Multi-Level Index Tables

Abstract

Simplicity is the major advantage of implementing hardware IP lookup engine using multi-level index tables. However, the memory efficiency of the conventional multi-level indexing approach is relatively low. In this paper we shall show that by restructuring the binary-trie using a method called bitshuffling, highly efficient index tables can be built to support the IP lookup operation. The proposed method is evaluated using a real-life IPv4 routing table with 321K prefixes. The amortized memory cost of our method is less than 3 bytes per prefix.

A 200Msps, 0.6W eDRAM-based search engine applying full-route capacity dedicated FIB application

“…Algorithmic search methods without TCAM have also proposed as considerable interest, which comprise of three major categories: tree-based methods [4], range search methods [5,6], and hashing-based methods [7]. TCAM and number of algorithm based can search against the lookup table with wire speed rate.…”