Packet classification is primarily used by network devices, such as routers and firewalls, to do additional processing for a specific subset of packets. Such additional processing includes packet filtering, quality of service (QoS), and differentiated services (DiffServ). Most of the existing packet classification algorithms reported in the literature exploits the characteristics of filtering or classifier rules in optimizing their techniques. However, the seminal observation made by Gupta and McKeown that a given packet matches only few rules in the classifier shows promise to another direction that packet classifier's average performance can be improved by exploiting the locality in the incoming traffic pattern. In this paper, we undertook the investigation of finding the feasibility of exploiting the locality in traffic to improve packet classifier's average performance. Our lightweight traffic-aware packet classifier reorganizes its internal data structure (rule tree) based on the traffic pattern to reduce the search time for the most frequently visited rules in the rule-set. Unlike existing traffic adaptive packet classifier requiring a separate, offline reorganization phase, our approach performs reorganization online with little overhead, resulting in higher throughput without compromising the accuracy. Experimental results on our test bed show that our traffic adaptive packet classifier incurs small number of memory access (i.e. less time per packet) in order to classify the packet.
About the presenter:
Min Sik Kim is an Assistant Professor of Computer Science at Washington State University. Prior to joining Washington State University, he co-founded and served as Chief Technology Officer of Infnis, Inc., a network security company. Prof. Kim's research interests include overlay networks, network monitoring, and network traffic analysis. His recent research focuses on the construction and maintenance of an efficient overlay network topology through measuring and analyzing network performance. Prof. Kim earned a B.S. degree in Computer Engineering from Seoul National University in 1996, and a Ph.D. degree in Computer Science from the University of Texas at Austin in 2005. He is a recipient of the Microelectronics and Computer Development Fellowship from the University of Texas at Austin.