Scalable Multi-Field Packet Classification Using Fixed-Length Compression Bit Vector

Packet classification is one of the core techniques in network devices such as firewall, IDS and IPS. Tens of thousands of rules pose great pressure on classification algorithm in terms of memory requirements and throughput. In this paper, a new packet classification algorithm, named Fixed-Length Compression Bit Vector (FLCBV), is proposed based on the compression scheme FLC. Compared with traditional algorithms, FLC is more adaptive to network environment, especially in large scale ruleset and multi-field packet header. Experimental results demonstrate that FLCBV takes only 65% of memory compared with traditional compression algorithms without losing much throughput.