Research On Controller Load Balancing And Data Center Network Traffic Scheduling Based On SDN

With the rapid development of Internet technology,emerging technologies such as cloud computing and virtualization have put forward higher requirements for network management and configuration,traditional closed network architecture can no longer meet the application requirements.Software defined network(SDN)is a new type of network architecture that decouples data forwarding and control decisions.It can centrally manage and control the network through the controller,and has the advantages of flexible programmability and easy deployment of new applications.SDN has accelerated the development and innovation of the network,and has been widely used in data centers and enterprise networks.It has become a hot topic in current research,but the development and application of SDN technology is still facing many challenges.On the one hand,the multi-controller architecture must have problems such as load imbalance and controller failure,implementing an effective controller dynamic load balancing algorithm and failure handling mechanism is essential to improve the performance and reliability of the control plane.On the other hand,because the traffic of internal communication in the data center network is complex,diverse and difficult to manage,which is easy to cause network congestion.Therefore,it is necessary to implement effective traffic scheduling strategy with the advantages of centralized management and programmability of SDN.In view of the above problems,the following two parts are mainly researched.(1)In order to solve the problems of controller failure and load imbalance in SDN multi-controller architecture.This thesis proposes a SDN-based controller failure handling(CFH)mechanism and a multi-constrained load balancing(MCLB)algorithm,and a dynamic decision controller architecture is designed.Each controller automatically selects the decision controller based on the collected load information and transfer conditions,then decision controller collects the global load information,per-forms the load balancing decision and completes failure processing tasks.The CFH mechanism provides two ways to deal with controller faults,when the ordinary controller fails,the decision controller will migrate the switch under the failure controller according to the failure handling mechanism,when the decision controller fails,the other controllers will reselect a decision controller based on the designed failure confirmation mechanism and complete the switch migration decision.When load imbalance occurs in the control plane,the MCLB algorithm defines multiple constraints to select the migration controller,target controller,and migration switch based on the controller's equalization degree and load migration efficiency,and uses the migration conditions to constrains the migration process,so as to improve load migration efficiency.The main idea is to make the controller's load closer to the average value of the global load.In order to prevent the loss of flow request messages during the process of switch migration,this thesis also uses a smooth switch migration method.The experimental results show that the proposed scheme in this thesis is superior to the relevant algorithms in improving the load migration efficiency and load balances degree of the control plane,and has the ability to cope with controller failure,ensuring the load balance and reliability of the control plane.(2)Aiming at the shortcomings of traffic scheduling selection algorithm in data center network.This thesis proposes a large-flow scheduling selection mechanism based on congestion probability(LFSSCP)in data center network by using SDN technology.The controller uses the OpenFlow protocol to obtain network status information,and performs functions such as calculating link load,identifying large flows and monitoring the congestion degree of the link.Firstly,LFSSCP calculates the equal-cost shortest path for each large-flow on the congested link and removes the path that does not meet the flow transmission bandwidth requirement,Then,the available throughput of each remaining path is calculated,and the path with the largest available throughput is taken as the optimal scheduling path,Finally,the congestion probability of each large-flow is evaluated by considering the transmission band-width and optimal scheduling path.The main idea of LFSSCP is to select the optimal scheduling object according to the congestion probability.In order to minimize the load on the scheduling path while reducing the congestion link load,and ensuring the load balance of the traffic on the link.During traffic scheduling,a reverse flow table update strategy is also provided to prevent flow interruption.The experimental results demonstrate that the mechanism can effectively improve the service quality of traffic transmission and achieve a better flow scheduling selection strategy for data center network.