The High Speed Packet Classification Supporting Multi - Field for Flow Tables in OpenFlow

The OpenFlow protocol specifies a variety of match fields, including exact fields, prefix fields, and range fields. Existing hardware-implemented packet classification causes prefix expansion when searching for a range field, resulting in large memory requirement. In order to reduce the cost and energy consumption, the high speed packet classification supporting multi-field for flow tables in OpenFlow is proposed. The solution is based on bit-vector method. Firstly, divide fields into two groups according to the matching type of fields. Then, select suitable matching methods for each match field to get intermediate results. Finally, a bit-wise logical AND operation among all intermediate results is performed to get the final matching result. The hardware implementation structure is a two-dimensional pipeline structure. According to theoretical analysis and simulation results, the solution can sustain a throughput of 100 Gigabit Per Second (Gbps).