Research On Reliability Guarantee Mechanisms In Software Defined Networks

The rapid development of Internet technologies(e.g.,Internet of things,cloud com-puting and mobile computing)has resulted in an exponential growth of Internet devices as well as the network traffic.At the same time,the devices need to support more proto-cols.All of these facts make the Internet architecture become more and more complex,which also increases the device complexity and the difficulty of network management.In recent years,to overcome these barriers,the Software-Defined Networking(SDN)has been provided.In SDNs,the control plane and the data plane of the network are decoupled into independent devices to realize the convenient management and flexible control of the network.With the advantage of centralized control and network pro-grammable,SDN provides a novel solution for network management problems such as network traffic monitoring,dynamic routing and network function virtualization.For these reasons,it has attracted wide attention in the academic and industrial circles.With the promotion and development of SDN,the reliability of SDN has been paid more and more attention.According to the existing research results,SDN network reli-ability is mainly manifested in three aspects:the strategy execution reliability,network connectivity reliability and network performance reliability.The network performance mainly includes the performance in control plane as well as data plane.As an emerging network technology,SDN has many shortcomings in reliability.First of all,due to a va-riety of network applications share the flow table resource in the data plane,applications with different control logics will lead to a conflict in flow control,which will reduce the reliability of network strategy execution.Second,in the multi-controller architecture,the expanding of network size and burts of network traffic can easily lead to high loads and high response delays in certain controllers,thus reducing the performance reliabil-ity of the control plane.In addition,the current SDN switches have serious bottlenecks on the flow capacity and control link capacity,leading to a negative effect on routing deployment delay and network transmission performance.Therefore,it brings a new challenge for data plane performance reliability in SDN.To address these issues,this dissertation will study the reliability guarantee mechanisms in SDN in the aspects of strategy execution reliability,control plane performance reliability and data plane per-formance reliability.The main research contents and contributions of this thesis are as follows.1.The architecture of software defined network,which separates the control plane and data plane,brings the advantages of network programmability and centralized man-agement.However,managing SDN flow rules,especially among multiple modules and administrators,has become complex and error-prone.Different controller modules with diverse objectives may be installed on the SDN controller,which can lead to anoma-lies among policies and rules.Aiming at the problem of strategy execution reliability,we first analyze the potential rule anomalies according to the flow entry definition in the OpenFlow protocol,and formalize the rules relationship and rule anomaly.On this basis,we present a rule anomaly detection algorithm based on state transition graph to realize the anomaly detection between two rules.To improve the performance of al-gorithm,we design a model of interval tree to store of the existing rule set,in order to reduce the scope of anomaly detection and improve the anomaly detection rate.Finally,through the deployment test on the OpenDaylight platform,we verify that our algorithm can effectively detect and resolve rule anomalies in SDN.2.With the expansion of network scale,the software defined network often re-quires multiple controllers to share the computing load in the control plane and avoid the single point failure.In the multi-controller scenario,how to reduce the response delay of the controller is a key problem to improve network reliability.To achieve this goal,the existing researches mainly include two solutions:multi-controller static de-ployment mechanism and multi-controller dynamic allocation mechanism.However,both of these approaches can lead to additional control layer communication overhead.In this dissertation,we will study the minimum controller response time problem based on flow redirection method.We formalize the problem by jointly considering the flow table size and data link capacity constraints and prove its NP hardness.On this basis,we propose two approximate algorithms to solve this problem.The simulation experiment and platform test show that the approximate algorithm proposed in this study can reduce the maximum controller response delay by 50%-80%compared with the two existing mechanisms.Compared with the dynamic allocation mechanism,the algorithm can reduce the number of controller deployments by nearly 30%with the same controller response time.3.In SDN,the data plane and the control plane are separated,and all traffic routing decisions in the data plane need to be made by the control plane.In order to improve the reliability of data transmission in the software defined network,most of the pre-vious researches only considered the constraints of flow entry resources and data link capacity in the data plane.However,due to the limited processing power of Open-Flow switch,the control link between switch and controller also becomes a bottleneck of network reliability guarantee in SDN.When burst flows occur in the network,the constraint of the control link will degrade the network QoS performance(for exam-ple,network throughput,transmission delay,etc.).In this thesis,in order to improve the reliability of the network and the network quality of service,we will focus on joint resource optimization by taking the data plane resource constraints and control link re-source constraints into consideration.For the uplink capacity constraint,we study a control link load balancing problem minimize the maximum control load and give an algorithm with(3 + 2(?))-approximation for solving this problem.For the downlink capacity constraint,we propose a low latency guaranteed routing algorithm to optimize the routing performance with route deployment delay constraint.Finally,through plat-form experiment and simulation experiment,we verify that the proposed algorithm can reduce the control link load by nearly 50%and reduce the route deployment delay by 60%compared with the previous methods.