Baking the ruleset: A heat propagation relaxation to packet classification

As one of the most critical components in network appliances, the packet classification method has to deal with two frequently contradicting requirements: to classify the packets at line speed and to update the ruleset quickly. Tuple Space Search (TSS), a classical hash-based packet classification algorithm, achieves fast rule updating at the sacrifice of the packet classification rate. In TSS, each tuple is managed by a hash table and classifying a packet needs to go through all hash tables. Merging tuples can reduce the number of hash tables, but the improper merging scheme will increase the hash collisions that may even worsen the classification performance in some cases. In this paper, we propose a novel packet classification scheme to achieve fast packet classification and online rule update simultaneously. By using heat propagation to relax the tuple merging optimization problem, our method can reduce the number of hash tables while keeping the number of collisions low. Experimental results demonstrate that our method achieves 3.2× classification speed and 4.6× update speed on average compared with state-of-the-art algorithms.