Research And Implementation Of Failure Detection And Recovery Techniques In SDN Network

With the rapid expansion of the Internet applications, types of services become richer and richer,and Internet applications make increasing demands on various functionalities of Internet. Software Defined Networking(SDN) is a new programmable network architecture, which separates the network control plane from data plane, and can satisfy various requirements of the Internet applications. Detecting and recovering failure in time is a very important criteria for evaluating the practical value of network, therefore, it is necessary to study the failure detection and recovery techniques in the SDN network to shorten the time of service disruption.This thesis discusses the structural characteristics of SDN and protocol features of Open Flow v1.3, and generalizes the failure types currently existing in SDN network and the detection method of data plane failures. By analyzing the features of the existing two SDN failure recovery mechanisms and their limitations, a synthesized failure recovery approach combined with proactive recovery and reactive recovery schemes, including that of considers the quality of service of packet flow, are proposed. A prototype system based on open source Open Daylight controller are designed and implemented, which uses asynchronous alert to detect network failures, and diagnoses failures depending on the number of failures and the service properties of flows affected by the failures. Three recovery modes, includes proactive, deliberatively proactive, reactive modes, are proposed to recovery automatically according to the diagnostic results.The function and performance tests are performed on the implemented prototype system of the failure detection and recovery by simulating SDN failures using Mininet. The test results show that the prototype system can detect network failure, and automatically uses corresponding failure recovery mode based on failure conditions.