GreenTCAM: A memory- and energy-efficient TCAM-based packet classification

Abstract

Ternary Content Addressable Memory (TCAM) is the de-facto standard device used for high-speed packet classification. Despite its capability for line-speed queries, it is very power hungry and area inefficient The latest TCAM devices by leading vendors come with an power saving mechanism where a subset of its TCAM blocks can be selectively activated. Recent research efforts exploit this feature to reduce power consumption with pre-classification steps. However, the state-of-the-art technique achieves power savings at the expense of poor utilization of TCAM capacity, and the potential of power reduction is not fully exploited in many cases. In this paper, we propose GreenTCAM, an optimized two-stage design for TCAM-based packet classification. Based on common characteristics of rule sets, our design is able to group rules more compactly into TCAM blocks, and activates a minimum subset of these blocks for each incoming packet. Experimental results show that our design achieves a 93.6% power reduction with a TCAM storage overhead of only 5.6% on average.