Abstract: Network big data collection is the basis of network behavior research, which can provide real and effective data basis for understanding the characteristics of network behavior. The existing FlowRadar measurement methods can encode and decode traffic for all streams, with less memory and constant count update time. However, each packet needs to be hashed, which has the problems of high memory access times and high computation cost. To solve this problem, this paper designs and implements the packet aggregation module and the flow evict module with P4 language based on Protocol Independent Switch Arch (PISA). Simulation results verify the optimization performance of the acceleration model in terms of throughput and network delay.

Key words: network big data collection, FlowRadar, aggregation-based acceleration