FlexTCAM: Beyond Memory Based TCAM Emulation on FPGAs

Abstract

Field Programmable Gate Arrays (FPGAs) have evolved into large system on chip devices that have high speed network interfaces, processors and vast amounts of logic and memory blocks. This coupled with their inherent reconfigurability, makes them an attractive platform to implement Software Defined Networking (SDN) systems. However, there is one piece missing that is used in most networking devices: Ternary Content Addressable Memories (TCAMs). TCAMs are widely used to implement packet classification. TCAMs can match an incoming key against a set of stored rules that include wildcard bits in one memory access. TCAMs are usually designed as specific hardware blocks and are included in almost all switching ASICs. However, they are not implemented in FPGA devices. This is due to the cost in terms of area/power of TCAMs hardware blocks and their lack of flexibility. To provide packet classification in FPGA based systems, a large number of schemes to emulate TCAMs on FPGAs have been proposed in recent years. In many cases, they are based on the use of the memory blocks available on the FPGA. In this paper, it is shown that FPGA emulated TCAMs are actually more powerful than traditional TCAMs and can support more generic rules. This makes them more efficient in some applications as the number of rules needed to implement a system can be significantly reduced. This is illustrated with two case studies: supporting port lists and ranges in access control lists and blacklisting of IP addresses. The results show the benefits of using more generic rules in the FPGA emulated TCAMs.