An Online Flow-Level Packet Classification Method on Multi-core Network Processor

Network packet classification is an important technique widely used in routers, firewalls and intrusion detection systems, and it plays an important role in network security. Many algorithms have been proposed to accelerate packet classification, such as various decision-tree based algorithms. However, each time a packet arrives, it is required to traverse decision tree to find matched rule in lead nodes, which is not efficient for online packet classification. Based on this observation, this paper proposes a flow-level packet classification method using flow table and decision tree, called FTDT. FTDT classifies packet according to the flow it belongs to, which is applicable to real-time packet classification. The proposed method is implemented in Tilera-gx36 multi-core network processor. Moreover, multi-thread technology is employed to design parallel packet classification scheme to accelerate packet classification. We tested our method with classbench rule set and campus rule set. Experiment results showed that FTDT had superior throughput compared with Hyper cuts.