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

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 bit-shuffling, highly efficient index tables to support the IP lookup operation can be built. The proposed method is evaluated using a real-life IPv4 routing table with 321K prefixes. The required lookup tables occupy 0.8MB on-chip memory. The memory cost is about 21 bits per prefix.