Efficient low power packet classification by grouping TCAM rules

Abstract

In this paper we propose a novel method for reducing power consumption in TCAM based packet classification. TCAMs are becoming very popular for designing high-throughput for packet classification. They are fast and simple to manage. However, due to their inherent parallel structure, a major drawback of TCAM is their high power consumption. Our method is based on using TCAM and its implementation on FPGA board. In this method, the packet classification process is done in two steps. At first, the rule set of the rule table is partitioned to some specific number of groups during an offline phase. Then, in real time phase, the decision process about each input packet is done based on only one of the partitioned groups. The group selection is done via the packet information, and the search operation will continue within that specific group. By partitioning the rules, storing each partition in a separate and smaller TCAM, and searching in only one TCAM, the power consumption will decrease which results to a more power efficient packet classification. The proposed method is implemented using FPGA. Implementation results show improvement in the power consumption.