Research And Implementation Of Packet Classification In IPv6 Networks

Packet classification is widely used in fields of Virtual Private Network, Distributed Firewall, Policy-based Routing, Differentiated Qualities of Service and Traffic Billing etc. For more IP addresses and higher quality of service, IPv4-based network is tending to be upgraded to IPv6 network gradually.Since this brings more pressures and new challenges to packet classification: 1) dealing with deeper protocol stack, 2) being compatible with IPv4, 3) best matching rule, 4) supporting content-based classification, this thesis mainly focuses on the research of: protocol analyzing, field extracting, rule managing and pattern matching. The attributions of the thesis include:1. Instructions for protocol analyzing and field extracting: An instruction-based hardware is independent on the depth of protocol stack. Compared with TCP/IP-based software, it processes packets at higher performance with less memory and supports new protocols dynamically.2. Distributed rules managing algorithm: Collisions often happen in multi-field classification. For packet to match the best rule, the thesis presents a distributed rules managing algorithm, which depends on the classes rather than the number of rules.3. Multi-Byte Shifting Multi-Pattern Matching algorithm: It is proposed for content-based classification by using ternary content addressable memory. It can support new patterns dynamically and multiple the processing bandwidths by enlarging the pattern space.The thesis presents a reference design of packet classifier that is implemented with FPGA and can be easily integrated to ASIC or NPU. It supports multi-field classification through layer 2 to layer 7 at 5Gbps wire speed.