Research On Distributed Controller Deployment Scheme Based On Software Defined Network

With the advent of the 5G era,the explosive growth of traffic in the network and the emergence of new network services,more and more drawbacks have appeared in the traditional network architecture.Therefore,the network architecture of SDN came into being.SDN network is efficient,flexible and easy to maintain,and many advantages make it become the mainstream of new network technology in the future.However,with the development of SDN technology,researchers found that the centralized controller structure is gradually becoming the bottleneck of SDN development.The centralized control method often brings about the problems of high delay and poor scalability,and there is also a very serious single point of failure hidden danger.Therefore,researchers began to propose various distributed controller structures,but at the same time they found that the number of controllers,the division of control domains,and the specific deployment location of controllers have a crucial impact on the performance of the entire SDN network.Therefore,proposing an excellent distributed controller deployment scheme has become a hot issue in SDN.This paper discusses the current research status at home and abroad,and proposes two solutions to the distributed controller deployment problem:(1)An improved particle swarm distributed controller deployment scheme(L-MPSO)based on community partitioning in SDN environment.The main body of the scheme is a two-step heuristic intelligent search algorithm based on community division.First,the network is initialized.The first step is to reuse the Louvain community detection algorithm to divide the network topology and obtain the number of controllers.The second step is to divide the control domain results.Enter the MPSO algorithm process as input,and output the optimal distributed controller deployment scheme after the iteration.The L-MPSO algorithm is compared with the LCPS algorithm and the RA algorithm.Through experiments under four different scale topologies,the delay,load balancing and reliability are evaluated respectively,and the effectiveness and advantages of L-MPSO are confirmed.(2)Hybrid Leapfrog Distributed Controller Deployment Scheme(L-SFLA)based on community partition in SDN environment.Considering that different intelligent algorithms often have different parameter settings and different search processes,the final search results will be different.And the SFLA algorithm is rarely used to solve the problem of distributed controller deployment.Another two-step intelligent search algorithm L-SFLA based on community partitioning is designed.Community partitioning and network initialization are consistent with L-MPSO,and the subsequent controller deployment position is determined by the SFLA algorithm.The algorithm is compared with the standard k-means algorithm and RA algorithm,and the topological BICS is selected for experiments.The algorithm mainly evaluates whether its search is effective,the weighted worst delay and the fault recovery mechanism,which proves that the performance of L-SFLA is better.This paper also conducts comparative experiments on the two proposed schemes.The conclusion is drawn that the L-MPSO scheme is more suitable for reducing the average delay and the network failure rate,and the L-SFLA scheme is more suitable for reducing the maximum network delay.