Implementing and evaluating a Self-Stabilizing Software Defined Network Control Plane

Abstract

Software Defined Networking (SDN) provides an attractive alternative to traditional, manual, error prone networks. SDN provides the opportunity to add programmability to networks, and decouple the control plane from the data plane. By doing so, a network administrator may control an entire network from a single SDN controller. While the work in this field is extensive, the question of how to maintain the connection between the control plane and the data plane has not received much attention. There is an algorithm called Renaissance that approaches the problem by utilizing self-stabilization, which is a strong notion of fault-tolerance for distributed systems. However, only a proof-of-concept implementation has been created for Renaissance this far. This report presents a prototype of the self-stabilizing SDN control plane algorithm that is implemented using the Floodlight controller. In addition we present an evaluation of the prototype. The evaluation shows that the prototype is fault-tolerant and able to recover from transient faults, which means that it is self-stabilizing. The stabilization time is shown to depend mostly on the degree of a network, but also on the number of nodes. Lastly, we present a proof-of-concept implementation using the ONOS controller.