Compact TCAM: Flow Entry Compaction in TCAM for Power Aware SDN

High throughput access to large data structures for lookup and classification have made Ternary Content Addressable Memory (TCAM) indispensable in today’s network switching devices. TCAMs offer single cycle lookup operation but at the expense of notoriously high power dissipation. While there is no suitable alternative to TCAM for maintaining line rate lookup, high power dissipation of TCAM have raised alarms in the growing power sensitive data centers. Moreover, these memories are expensive compared to other memory devices and have direct impact on cost of switching devices. The situation further worsens with the emerging network frameworks such as Software Defined Networks (SDN). SDN’s large sized flow entries (15 field tuple that requires 356 bits to define a flow) can potentially support large number of flows and consequently an SDN device will need large TCAM size to accommodate them. An SDN switching device would require 5 to 7 times the TCAM space to hold the same number of flow entries as a Layer-2 switching device. Therefore, as expected, high TCAM size in SDN enabled network switches will take a toll on power budget. In this paper we propose Compact TCAM, an approach that reduces the size of the flow entries thereby managing large sized and volume of SDN flows without adding significant silicon real estate in TCAM. We use shorter tags for identifying flows than the original number of bits used to store the flow entries. We in turn leverage the dynamically programming capability of SDN to route the packets using these tags. This reduces the power dissipation per flow. We show that our approach can be easily implemented using the new SDN framework while optimizing the TCAM space an impact of which can be realized in cost and power saving. Our experiments both with real world and synthetic traffic shows that the power can be reduced on a average by 80% in SDN switching devices for a given number of flows supported by an SDN switch.