Flow Monitoring in High-Speed Networks with 2D Hash Tables

Abstract

Flow monitoring is a required task for a variety of networking applications including fair scheduling and intrusion/anomaly detection. Existing flow monitoring techniques are implemented in software, which are insufficient for real-time monitoring in high-speed networks. In this paper, we present the design of a flow monitoring scheme based on two-dimensional hash tables. Taking advantage of FPGA technology, we exploit the use of parallelism in our implementation for both accuracy and performance. We present four techniques based on this two-dimensional hash table scheme. Using a simulation environment that processes packet traces, our implementation can find flow information within 8% of the actual value while achieving link speeds exceeding 60 Gbps for a workload with constant packet sizes of 40 bytes.