Distributed Control Plane Architecture for Network Elements

With the explosion in the use of the Internet for ebusiness, entertainment, and wireless communication, Internet control and routing infrastructure are facing limitations in terms of the increasing scale. Because the Internet is now mission-critical to many industries, providing high availability (HA) is a major concern. In this paper, we discuss the scaling limitations of the routing and signaling protocols used in the Internet and propose a novel method to distribute the functionality of control plane protocols implemented in network elements such as switches and IP routers. The proposed method relieves the scalability bottlenecks in various signaling and routing protocols by selectively off-loading certain periodic operations of the protocols to the data plane processing components within a network element. This leads to scalable, highly available, and robust signaling and routing protocol implementations. We use the popular routing protocol, Open Shortest Path First (OSPF) as an example to discuss the appropriate metrics for evaluating the performance of an OSPF implementation, and then we identify the bottlenecks in scaling-up an OSPF implementation. We also describe a generic method for distributing the functionality of routing and signaling protocols, such as OSPF, across both control and data planes that makes them (a) more scalable, (b) more resilient in the presence of faults, and (c) amenable to faster convergence in the face of link or node failures.