Compressed Data Structures for Recursive Flow Classification

High-speed packet classification is crucial to the implementation of several advanced network services and protocols; many QoS implementations, active networking platforms, and security devices (such as firewalls and intrusion-detection systems) require it. But performing classification on multiple fields, at the speed of modern networks, is known to be a difficult problem. The Recursive Flow Classification (RFC) algorithm described by Gupta and McKeown performs classification very quickly, but can require excessive storage when using thousands of rules. This paper studies a compressed representation for the tables used in RFC, trading some memory accesses for space. The compression’s efficiency can be improved by rearranging rows and columns of the tables. Finding a near-optimal rearrangement can be transformed into a Traveling Salesman Problem in which certain approximation algorithms can be used. Also, in evaluating the compressed representation of tables, we study the effects of choosing different reduction trees in RFC. We evaluate these methods using a real-world filter database with 159 rules. Results show a reduction in the size of the crossproduct tables by 61.6% in the median case; in some cases their size is reduced by 87% or more. Furthermore, experimental evidence suggests larger databases may be more compressible. Compressed Data Structures for Recursive Flow Classification Edward W. Spitznagel [email protected] Washington University St. Louis, MO 63130-4899 This work was supported by the National Science Foundation, ANI-9813723.