Research And Design Of The Improved Recursive Flow Classification Based On IXP2400

In modern computers network, there will be more kinds of application services and more broadband access. The networks equipment should not only forward the packets undistinguished, but also provided differentiated services for various users in the trend of multiple services. Packet Classification is the base technology for the differentiated services. Classifying packets on multiple dimensions is a difficult problem, the network equipment become the bottleneck in the trend of the broadband. So we should look for a better method with high speed.The RFC (Recursive Flow Classification) chosen by the paper has highest speed in the software methods after the analysis of the packets classification methods. The high query speed is the advantage of the RFC, and there are many distributions in the method, so parallel querying can implement in RFC. The high-speed method is implemented on the multi-core multi-thread Network Processor Unit (NPU) IXP2400.The disadvantage of RFC is excessive storage with increasing of the number of rules. So, the rule set compression and the crossproducting table compression is used in this paper. In the rule set compression, the rule set, in which the rule is the base element, is rearrange to a set based on the actions. The adjacency rules can be merged in the same-action-group. In each phase of the preprocessing of RFC, there are some fields combined, so rules compression can be achieved in each phase. The bit-victor compression is used in the crossproducting table compression to decrease the memory space belonging to the contiguous same elements. The data structure designing, flow chart designing, and the coding for the two compression schemes are achieved by this paper.On the NPU, the assignment of the processors and the thread is the key for the sufficient use the hardware resource and the line speed forwarding. The assignment for ME, threads, memory is illustrated in detail in the paper.The simulation is implemented on the NPU development platform. According to the simulation results, the storage space will increase excessively with the rule set increase. Also, the storage space has better improvement with increase of the rule set. In the best condition 54% compression rate can be achieved.