Clustering Boundary Cutting for Packet Classification Based on Distribution Density

摘要

In this paper, we present the clustering boundary cutting trie algorithm in order to solve the problem of huge time consumption in existing trie based algorithms. In the proposed solution, there are two stages. The first stage is the density-based rule clustering process. The rules are represented as a range between 0 and 1 according to the prefixes of the packet fields. When the number of the rules in a range reaches to a certain density, the corresponding rules are formed in a cluster. The second stage is the trie construction process based on these clusters. Compared with traditional packet classification algorithms, the searching time of our algorithm increases by 47.05%-73.76% and keep the high accuracy of 69.83%-93.17%. The experiment demonstrates that our algorithm can effectively keep high accuracy as well as keeping stable high-throughput, and it is suitable for actual deployment.