A Balanced Tuple Partitioning Method for Packet Classification with High-Performance and Scalability

The scalability of packet classification and dynamic rule updates in SDNINFV-driven networks remains a critical challenge as network sizes expand. To address this, we propose Balanced Tuple Partitioning (BTP), a novel two-phase optimization framework based on the Tuple Space Search (TSS) algorithm. BTP introduces (1) tuple merging-based partition, which strategically consolidates overlapping tuples to reduce rule-set complexity, and (2) conflict-aware tuple chaining, a lightweight mechanism to resolve post-merging conflicts while ensuring rule integrity. We formally prove the correctness of BTP through a analysis of tuple space efficiency (reducing search iterations) and rule space consistency (maintaining se-mantic equivalence). Experimental results demonstrate that BTP achieves 2.2x, 3.3x, and 1.8x higher classification throughput on average than DynamicTuple, TupleTree, and TupleChain, respectively. Furthermore, BTP sustains high performance under extreme scalability demands, supporting rule-sets of 10 million rules and dozens of match fields.