Discrete Bit Selection: Towards a Bit-Level Heuristic Framework for Multi-Dimensional Packet Classification

Packet classification is still a challenging problem in practice under large number of classification rules and constant growth of performance requirement. Most of the existing algorithms try to solve the problem heuristically by leveraging on the inherent field-level characteristics of the rules. This paper proposes a bit-level heuristic framework: discrete bit selection (DBS) for multi-dimensional packet classification. Preliminary experimental results show that DBS-based algorithm gains much better performance both in search time and memory requirement than the well-known field-level algorithms with various real-life rule sets.