A Systematic Non-Prefix Approach to Minimizing Packet Classifiers in TCAMs
Packet classification enables many networking services on the Internet, such as firewall packet filtering and traffic accounting. Using Ternary Content Addressable Memory (TCAM) chips to perform high-speed packet classification has become the de facto standard in industry. TCAM chips classify packets by comparing a packet with the full suite of classification rules in ternary encoding, the comparison being done in parallel.
Unfortunately, TCAMs have limitations of relatively small capacity, high power consumption & heat generation, and high cost. The well-known, range-expansion problem exacerbates these limitations by significantly decreasing the limited capacity of TCAMs as each classifier rule typically has to be converted into multiple TCAM rules.
Traditional packet classification looks at five fields: source and destination IP addresses, source and destination port numbers, and protocol type. New Internet services and new security threats are leading to more complex rule sets. Along with the increasing adoption of IPv6, the size and width growth of packet classifiers puts more demand on TCAM capacity.
Description of Technology
Michigan State University has developed a suite of algorithms to enable the optimal use of TCAM chips. This invention is another method that takes a given packet classifier as input and outputs a semantically equivalent packet classifier that requires fewer TCAM entries. It is a non-prefix classifier that uses two techniques, bit swapping and bit merging, to identify and merge the rules that can be combined. Bit Weaving can be used as an alternative to TCAM Razor or in combination with it.
Other inventions inventions in the suite include:
Patented US 8,375,165
Xiang-Yang Liu, Eric Torng, Chad Meiners
For Information, Contact:
Michigan State University