An Efficient Hashing-Based Scheme for RIB Management

In consideration of the exact matching characteristic of RIB lookups, we explore the applicability of hashing-based approaches for RIB management and evaluate their performance. Using real routing prefixes, we compare the efficiency of six different hash approaches mentioned in some papers about IP lookups: bit extraction from the routing prefix, CRC16 (cyclic redundancy checking polynomials), CRC32, Fletcher checksum, folding of prefix octets for using the exclusive-or operation and multiplication. This is the first comprehensive study of their performance using real core routing tables. We find that extracting the least significant bits from the routing prefix is a very efficient hash approach for RIB management. Then a RIB management scheme based on this hash approach is proposed. We use separate chaining as the collision resolution strategy, while not suffering from malloc/free overhead for frequent insertion or deletion. To keep the memory storage moderate, we set the hash table sizes based on the distribution of prefix lengths because it is not expected to change significantly with time. Through simulation, we find that this scheme can achieve very fast updates and fast lookups for routing prefixes while keep the memory storage moderate.