A tri-state-based partition approach to reducing power consumption of TCAMs

Ternary Content-Addressable Memories (TCAMs) have become the industrial standard for high-throughput packet classification. The comparison against all of its entries in parallel within a single memory access makes TCAMs power hungry. In this paper, we propose a new algorithm called TSP-PR (Tri-state-based Partition for Power Reduction). In TSP-PR, identification bits are dynamically selected to partition a classifier into one general subset and many dedicated subsets. When packet arrives, only one dedicated subset and the general subset are activated for parallel search. This makes a great power reduction since the main component of power consumption of TCAMs is proportional to the number of searched entries. The experimentation shows that the TSP-PR algorithm proposed in this paper can highly reduce power dissipation with a little penalty of storage overhead.