Research Of Multi-Controller Placement Based On SDN

Software Defined Networking (SDN) is an emerging networking paradigm that provides applications with an abstract centralized view of the overall network by separating the network control plane from the data forwarding plane. The idea of SDN has been widely accepted and the network architectures with decoupled control plane are now transiting from research prototypes to real deployments. The traditional SDN implementation relying on a logically centralized controller has several limitations related to scalability and performance. To deal with a large-scale WAN, where multiple controllers are required, the most fundamental task in the SDN context is the proper placement of the controllers.So far, researchers have already conducted a variety of related studies on multi-controller deployment issues in WAN, and proposed a variety of solutions. However, these existing solutions do not take the limited-capacity of controllers into consideration. On the other hand, the effects of controller failure on network performance are not too much considered. Therefore, combining the existing research achievements, we take a further step in the research of multi-controller deployment issues.Based on the analysis above, the main research contents of this paper are as follows:(1) In this paper, we propose a novel placement metric for deploying multiple controllers that measures the cost when controllers with limited capacity handle the request from switches. Then, we formulate the optimal controller placement problem as an Integer Linear Program and use an effective approximation algorithm to solve it. We conduct intensive experiments based on many real topologies and the results demonstrate that our strategy can significantly improve performance over existing ones in terms of both cost and load balance.(2) Considering the effects of controller failure on network performance, we introduce two metrics to study the multi-controller deployment issues in the case of some controllers may malfunction. In the corresponding optimization problem, the goal is to find one placement such that these two metrics is minimized in the case of at most a certain number of controllers malfunction. The experimental results show that there is not a placement reaching the lowest value on these two metrics simultaneously. For this reason, we need find the tradeoffs between these two metrics and then achieve Pareto-optimal placements.