| Summary: | 碩士 === 國立中興大學 === 資訊科學與工程學系 === 102 === Packet Classification is an important technique to enable a router to provide services, including firewall, virtual private network (VPN), quality-of-service (QoS), etc. It categorizes incoming packets into different forwarding classes corresponding to predefine rules. The performance of packet classification may influence the packet latency because of the extra delay for determine the matching rules. In this paper, we present a new algorithm using binary searches upon the multidimensional rules. We provide the necessary conditions that the multi-field rules in an array can be sorted to perform binary search. In addition, the algorithm supports incremental updates. We also use multi-way search trees and fractional cascading to further improve the search performance. The experimental results that the algorithm usually has superior speed and storage performance. Moreover, our scheme can support up to 100k rules.
|
|---|
