Research On Fault Recovery And Load Balancing Strategy Based On SDN Controller

With the development of emerging network technologies such as big data,cloud computing,and the internet of things,and the explosive growth of network scale,traditional network architectures have been unable to adapt to network technology innovation.Software-defined network(SDN)as a new network architecture with centralized control,open programmability,and easy deployment of network services,meets the needs of emerging network technology innovations.At present,SDN has become a hotspot of network technology research,and has been widely used in many scenarios such as data center networks.However,SDN-related technologies are still in the development stage,and there are still deficiencies in applications,and they face many challenges.On the one hand,the multi-controller architecture improves the scalability of the control plane,but the problem of unbalanced load among controllers and controller fault handling still affect the control plane's resource utilization and reliability.On the other hand,traditional load balancing algorithms cannot adapt to the diverse characteristics of network traffic in the data center,which can easily block the network and reduce network performance.With the help of SDN,it has the advantages of centralized management and control of the global network,thereby achieving effective management of traffic.In view of the above problems,the research in this dissertation is mainly carried out from the following two aspects:(1)For the load imbalance and controller reliability issues of the SDN control plane,this dissertation proposes a control plane load balancing and failure recovery(LBFR)mechanism based on dynamic controller allocation,and implements a centralized decision controller dynamic allocation architecture.In this dissertation,controllers are divided into decision and ordinary types.Decision controllers obtain the load and status information of ordinary controllers,and complete load-balancing decisions and controller failure handling based on the global network status.The status monitoring module periodically checks the network status.If the load among controllers is unbalanced,the LBFR achieve a load balancing migration strategy to select the controllers and switches to be migrated and the target controller to be immigrated based on the controller's load,controller processing capacity,and communication cost.If ordinary controller breaks down,the fault handling module will consider factors such as delay and load to reassign controllers to switches in the control domain of the faulty controller.For a decision controller break down,the ordinary controller sets a backup decision controller as the decision controller ahead,and the new decision controller accomplishes the fault handling task.Finally,the switch configuration module accomplishes the switch migration strategy generated by the controller load balancing module and the fault handling module to achieve the purpose of controller load balancing and fault recovery.The experimental results show that the algorithm proposed in this dissertation is superior to other algorithms in load balancing among controllers,and can effectively handle controller failures,thereby improving the load balancing capability of the SDN multi-controller architecture and ensuring the reliability of network functions.(2)For the problems that the traditional Equal-Cost Multi-Path routing algorithm(ECMP)in the data center network cannot effectively schedule large flows,causing the traffic load to be unbalanced and network congestion,Ant-colony algorithm based on Dynamic Multipath Load Balancing(ADMLB)is proposed in this dissertation.The ADMLB algorithm first monitors the load of data plane and topology information through the southbound interface of the controller,and the end host detects and marks the elephant flows based on the size of the data flow.Then an improved ant colony algorithm is called according to the influencing factors such as the data plane link bandwidth,delay and switch cache.Afterwards the optimal multi-path distribution of the elephant flows is selected based on the bandwidth and load to avoid link congestion.The experimental results show that the ADMLB effectively improves the link bandwidth utilization and reduces the link delay time compared with the traditional ECMP algorithm and the existing related flow scheduling algorithms.