A Survey of Packet Classification Tools and Techniques

Abstract

Network processors are evolving to meet the increased bandwidth demands placed on computer networks. Packet classification is one of the significant tasks performed by Network processors in this regard. It is accomplished through various algorithmic techniques and is implemented on different kinds of hardware. The work presented here analyzes the state-of art packet classification techniques based on these algorithms and hardware platforms. Implementation of classification algorithms based on clustering, bit vectors and trees on different hardware platforms are studied and performance is compared in terms of scalability, throughput and memory utilization. The survey indicates that packet classification techniques start from 5 fields and scale up to 15 fields for performing classification. It is also observed that the multi-core implementation of scalable packet classification technique utilizes a maximum memory footprint of 32Kb.