Title: Flow Correlator: A Flow Table Cache Management Strategy

Abstract: Switching, routing, and security functions are the backbone of packet processing networks. Fast and efficient processing of packets requires maintaining the state of a large number of transient network connections. In particular, modern stateful firewalls, security monitoring devices, and software-defined networking (SDN) programmable dataplanes require maintaining stateful flow tables. These flow tables often grow much larger than can be expected to fit within on-chip memory, requiring a managed caching layer to maintain performance. This paper focuses on improving the efficiency of caching, an important architectural component of the packet processing data planes. We present a novel predictive approach to network flow table cache management. Our approach leverages a Hashed Perceptron binary classifier as well as an iterative approach to feature selection and ranking to improve the reliability and performance of the data plane caches. We validate the efficiency of the proposed techniques through extensive experimentation using real-world data sets. Our numerical results demonstrate that our techniques improve the reliability and performance of flow-centric packet processing architectures.