A Hybrid Asymmetric Traffic Classifier for Deep Packet Inspection Systems with Route Asymmetry

Abstract

A flow is said to be asymmetrically routed if its packets follow separate paths for forward and reverse directions. Routing asymmetry leads to problems in flow identification, policy enforcement, quota management, traffic shaping etc. in DPI systems. There are two existing approaches to battle routing asymmetry: clustering and state sharing. The latter fails with stateless traffic, while clustering leads to large traffic overhead. We propose the Hybrid Asymmetric Traffic Classifier (HATC) method that merges the best aspects of the two existing methods. HATC is able to handle all types of asymmetric traffic with reduced overhead compared to clustering. Numerical evaluation of HATC using two real traffic traces is also presented.