An Efficient CPP Solution for Resilience-oriented SDN Controller Deployment t

摘要

Software-Defined Networking (SDN), as an emerging network paradigm, has attracted considerable attention from both industry and academia. One of the most intriguing study directions is the capability of fast recovery in SDN, where several works have been using multiple controllers to address the failure of a centralized controller. However, the limitation of such approaches is the inefficiency of exclusive controller processing, which provokes resource idling or suffers sudden flow outburst. In this paper, we propose ReON - an efficient optimization method for resilience associated with Network Function Virtualization (NFV) technique, deploying the controller and switch functions into one machine. Thereby, the device is capable of executing the respective functions according to the role it presents. This work formulates the controller resilience issue as the Controller Placement Problem (CPP), decentralizing the resilient ability without destructing the properties of SDN. Meanwhile, we also investigate applying Simulated Annealing (SA) in the proposed scenarios and compare the performance among the proposed method, SA and CPP brute force. Experimental results show the proposed scheme reduces the time complexity while keeping 95% accuracy against with baseline.