SPARC: Towards a Scalable Distributed Control Plane Architecture for Protocol-Oblivious SDN Networks

摘要

High-level programming abstraction and large-scale deployment have become two important trends of software-defined networking (SDN) in the past decade. Using high-level program to manage the large-scale network faces more serious control plane extensibility problem due to its complex intermediate representation and protocol-independent feature. To address this problem, we propose SPARC, a programmable and scalable controller architecture, which employs a hybrid hierarchical structure to maximize flexibility with regard to control plane distribution. Our architecture also allows for pushing down control decision making closer to the data plane and localize network event processing to lower the latency of control plane operations while exploiting SDN's global visibility to build optimal policy decisions. Furthermore, we investigate the feasibility of SPARC by exemplifying the case of delivering ICN mobility services and then conduct evaluations to demonstrate the efficacy of our design.