PartitionFilter: A Fast Packet Classification Algorithm Based on Adaptive Partition and Filter

Zhuo Li, Yanli Pei, Hao Xun, Jindian Liu, Peng Luo, Yu Zhang, Teng Liang, Kaihua Liu

Abstract

The decision tree, as an efficient data structure, is commonly used to perform packet classification due to its faster classification speed. During packet classification, all packets need to traverse each decision tree in descending order starting from the highest priority to match the rules. Therefore, the lookup performance of the decision tree with the highest priority has an important impact on the total classification performance. Moreover, reducing the number of decision trees to be accessed is an effective way to improve the classification performance further. In this regard, a new scheme called PartitionFilter is proposed in this paper. PartitionFilter adopts an adaptive rule set partition method (RSP) to reduce the lookup time of the decision trees (not only the highest-priority decision trees). A decision tree filtration method (DTF) is also proposed to reduce the number of traversals among decision trees. Experimental results show that the throughput of PartitionFilter is improved by an average of 54%, and the number of memory accesses is reduced by an average of 76% compared to PTTree.
BibTeX
@article{li2026partitionfilter,
  author = {Zhuo Li and Yanli Pei and Hao Xun and Jindian Liu and Peng Luo and Yu Zhang and Teng Liang and Kaihua Liu},
  title = {{PartitionFilter: A Fast Packet Classification Algorithm Based on Adaptive Partition and Filter}},
  journal = {IEEE Trans. Network Science \& Engineering},
  year = {2026},
  doi = {10.1109/tnse.2025.3582928},
}

Loading related content...

Figure