Computer Science

Scalable and balanced policy enforcement through hybrid SDN-label switching

Olufemi Odegbile, Herbert Wertheim College of Engineering
Shigang Chen, Herbert Wertheim College of Engineering
Youlin Zhang, Herbert Wertheim College of Engineering

Abstract

Software-defined networks facilitate automatic poli- cy enforcement with dynamic routing of flows through a sequence of middleboxes that offer the required network functions. As a result, network policy enforcement based on middleboxes, which is tedious and error-prone to perform in traditional IP networks, is greatly simplified. However, TCAM-based flow tables in SDN are small and energy-demanding, which limits the scalability of policy enforcement. This paper proposes a hybrid SDN-label switching scheme that combines TCAM- based switching (in SDN) at the network edge with label switching in the network core to provide scalable policy enforcement without compromising per-flow management capability. A linear optimization is proposed to balance workloads among the middleboxes. We demonstrate on OMNET++ that our proposed solution incurs much smaller processing/communication overhead and achieves better load-balancing when comparing with the prior art.