Research On The Optimization Of Distributed Control Plane Placement In SDN

The software defined network(SDN)advocates a centralized network control,where a controller manages a network from a global view of the network.However,in a large-scale WAN deployment,this rudimentary centralized approach with single controller has several limitations related to performance and scalability.So the study of distributed control plane has drawn the attention of researchers,in which the controller placement problem has become a researching hot area.This paper focus on the research and application of controller placement problem in SDN distributed control plane.We pointed out the controller architecture development trends from static architecture to dynamic architecture.Multi-controllers placement problem can be summarized as follows.In a given network topology,how many controllers we need to meet network requirements.Which method is optimal to find out the optimal location for the deployment of controllers and to realize dynamic optimization of control plane subject to the variation in traffic? This paper includes following work and contributions:1)Aimed at controller number selection and its effect to control path in controller placement,we proposed a method to realize controller placement based on control path connectivity.First,we established failure scenario metric model and defined control path connectivity.Controller placement problem could be transformed into the problem of searching the minimum failure scenario metric.Second,a controller placement algorithm based on greedy(GBCP)was designed to complete the partition of controller domain.Theoretical analyses and experimental results show,control path failure rate could be reduced and system reliability could be improved by the regulation of controller numbers in network topology.Meanwhile,this method could fulfill latency constraint.2)Aimed at control overhead problem after the work of controller placement,we proposed a controller placement scheme based on node cost and control path connectivity efficiency.First,we defined node cost,which included data connection cost,cost of route request and route installation and synchronization cost.Next,we used Top-k strategy established in deployment model to find optimal controller placement location in the node set,which meet control path connectivity threshold.Theoretical analyses and experimental results show,rational selection of three metric in node cost could reduce system control overhead after controller deployment.3)In existing research,control plane could not have comprehensive consideration to reply the changing load distribution between the controllers and network dynamic variation.Aimed at the problem,we put forward switch migration algorithm(SMA)and controller pool scalable capacity algorithm(CSCA).Switch migration algorithm was used to migrate switches in overloaded controller domain to the under-load controller domains,which could achieve the load balancing among controllers.Controller pool scalable capacity algorithm was used to grow or shrink the controller pool to adjust controller numbers in the pool and improve system utilization efficiency.Theoretical analyses and experimental results show,in two different random topologies,active controller numbers and system management overhead in this scheme are less than the other comparison schemes.