Multidimensional packet classification with improved cutting

Packet classification based on decision tree are easy to implement and widely employed in high-speed packet classification. The basic objective of building a decision tree is minimal storage and time complexity. HyperIC is a multiple dimensional packet classification algorithm. It is an improved HyperCuts algorithm based on statistics and evaluation on filter sets. The proposed algorithm allows the tradeoff between storage and throughput during creating decision tree. It is suitable for IPv6 packet classification as well as IPv4 because it is not sensitive to length of IP address. The algorithm applies a natural and performance-estimated decision-making process. We define maximum storage occupied and then achieve the best throughput. Evaluation shows that HyperIC provides a great improvement over HiCuts and HyperCuts algorithm in both storage requirement and searching performance and scalable to large filter sets.