摘要
Packet classification, which classifies the incoming packets according to their head information, is an important technique for the next-generation routers, firewalls and so on. Recursive flow classification algorithm is one of the fastest software packet classification algorithms, but its initialization time is too long, so a timesaving recursive flow classification algorithm is proposed. In phase 0, equivalence classes are defined by the set of filters matched in processing the packet. A more effective method for finding every entry's equivalence class is given. In phase 1, in order to combine the results for different chunks together, cross-producting tables are used to store precomputed results and a faster method of merging two chunks is given. Its initialization time is greatly reduced because of heuristic learning while its classification phase retaining the same time complexity with RFC. The experimental results show that the total time has an average decrease of 40%.
- 出版日期2009
- 单位南京航空航天大学