Packet Classification with Explicit Coordination

Packet classification is a key building block of many network services and functionalities (e.g., switching, filtering, load balancing). Despite its prevalence, packet classification is implemented and deployed in an ad-hoc manner at different layers of the protocol stack. Moreover, high speed packet classification, in presence of arbitrarily large number of classification rules, is resource and computation intensive. We argue that instead of developing classification solutions in isolation, packet classification should be considered as a fundamental primitive in the protocol stack. We propose a classification layer or CLayer as part of the protocol stack and a generic mechanism to explicitly configure and implement capability-driven classification offloading. Our approach allows classifiers (e.g., routers, firewalls) to offload part of their classification related computation and memory requirements to helpers (e.g., end hosts, edge routers) without introducing additional states in the classifiers. Evaluation results from our prototype implementations attest CLayer’s scalability and show increasing performance gain as traditional classification complexity grows.