Research On Optimization Algorithms Of Load Balancing Among Controllers Based On Latency Constraint In Software Defined Networks

Software Defined Networks(SDN)that separates the control plane from the data plane by Openflow has allocated control logic to the centralized controller.In large-scale WAN,existing single-controller placement strategies have some flaws in bandwidth and latency,etc.,therefore,the multi-controller placement problem has been emerging.The facts that how to determine the optimal number of controllers required for the network and how to place multiple controllers at appropriate locations are the key of the controller placement problem.The research shows that the efficient controller placement strategy can not only ensure the load balancing among controllers,but also reduce the network latency.In this thesis,the main contents of optimization algorithms of load balancing among controllers based on latency constraint in Software Defined Networks are as follows.(1)the research on optimization algorithm of load balancing among controllers based on latency constraint in initial static network:For the purpose of solving the problem that network requests can't be processed in time due to the load imbalance among controllers and the excessive network latency,the initial static placement strategy proposed in this thesis is as follows.According to the network traffic,the maximum processing capacity of a single controller and the decay factor,the number of controllers(M)required for the initial network is determined firstly.Then,a greedy strategy is used to determine the position of first controller,and constructing the first improved minimum spanning subtree with limited load based on first controller which is as a root node.Furtherly,an improved greedy strategy which is distinct from above is used to determine the positions of remaining M-1 controllers,and constructing the remaining M-1 improved minimum spanning subtrees with limited load respectively based on M-1 controllers which are as root nodes.The initial set of controllers is updated for minimizing propagation delay lastly,therefore,the initial M control domains are generated.Simulation results show that as to compare to PSA algorithm,optimized K-Means algorithm,etc.,the proposed initial static placement strategy can make network load balancing index increase by 29.65% averagely and guarantee the lower propagation delay.(2)the research on adjustment algorithms of load balancing among controllers based on latency constraint in dynamic network:Due to the network traffic changes with time,so the research is studied furtherly on the basis of the above research and adjustments of control domains in dynamic network are divided into three situations:1.The total traffic of network is not overloaded,an online optimization algorithm is proposed to adjust the relationships between controllers and switches so that each controller can work within efficiency range,simultaneously,it optimizes the network latency.2.The total traffic of network is overloaded,a controller activation algorithm is proposed to determine the number and locations of new activation controllers,then,the optimization algorithm in situation 1 is used to adjust the relationships between controllers and switches.3.The total traffic of network is under the lower limit of efficiency range,a controller dormancy algorithm is proposed to determine the number and locations of controllers in need of dormancy,then the optimization algorithm in situation 1 is used to adjust the relationships between controllers and switches.Simulation results show that the above three adjustment algorithms can be applied to three dynamic network situations respectively,on the basis of ensuring the network connectivity,they can make network load balancing index increase by 34% averagely and guarantee the lower propagation delay.