Research On Traffic Scheduling And Fault Recovery In Hybrid SDN Networks

The Internet has become the information infrastructure of human society.With the increasingly market competition of Internet service providers and the rapid increase of Internet applications,the Internet has already become an indispensable part of the modern information technology.The stability and efficiency of the Internet is very important to the national strategy.Nowadays,the Internet mainly uses OSPF/IS-IS routing in the intra-domain and BGP routing in the inter-domain.When the link load is congestion,or the network failure causes packets loss,the network equipment cannot guarantee stable quality of service and efficient network utilization.Therefore,how to achieve stable and efficient network services,how to ensure the quality of service when network failures,how to improve network utilization and how to use network services more economically and environmentally,that have become the important issues to the academia and industry.This paper mainly studies how to incrementally deploy Software Defined Network(SDN)switches with low cost in the networks,and how to realize network requirements such as traffic engineering,fault recovery and energy saving using SDN switches.The existing network protocols have the following shortcomings: 1)With the development of SDN network,the deployment cost is very high and SDN switch upgrading is troublesome.Ternary content addressable memory(TCAM)flow table entries of SDN switches are not enough,and it is difficult to install and update the flow tables.2)Network link failure and node failure require a lot of manpower and material resources to configure the network.Network failure is not only difficult to detect but also time-consuming to recover.The traffic of data center is uneven,and the link utilization is low.3)With the development of data center network,the energy consumption of data center network is increasing.But the energy utilization rate is very low,so the efficient energy-saving scheme is needed.4)In recent years,more and more efforts have been made to optimize the scheduling based on a group of semantically related flows(also called a Coflow)as the basic unit in the data center.However,the impact of the available bandwidth changing when scheduling has been ignored,so a new real-time Coflow scheduling algorithm is needed.To solve the above problems,this paper proposes an efficient solution by incremental deployment of SDN switches.The main contents include:1)This paper proposes an incremental SDN switch deployment scheme based on hybrid SDN network,which maximizes network control capability while minimizing the cost of SDN switch deployment.2)Aiming at the difficulties of network fault recovery,complex flow table configuration and long recovery time,this paper design a traffic scheduling and fault recovery scheme based on hybrid hierarchical SDN network.The scheme supports the preconfiguration of routing function and the controller temporary fault.With deploying a few SDN switches,the scheme can achieve 100% single link fault recovery.3)Aiming at the problem of low utilization of flow table entries in SDN resource management,this paper proposes a new low-load flow control algorithm and flow table entries allocation scheme,which is a data-layer forwarding mechanism based on MPLS tag.The scheme can effectively achieve the balance of flow table entries between SDN switches while reducing the number of flow table entries installed.So that the network can accommodate more flows.4)Aiming at the problems of high energy consumption of network equipments and insufficient utilization of network bandwidth,this paper proposes a energy-saving scheme based on SDN switch deployment to solve two different scenarios:(1)Give a limited number of SDN switches,and how to deploy them to achieve maximum network data flow control capability.(2)Fixed the location of SDN switch deployment,how to maximize network energy saving by re-routing the flows.5)Aiming at the problem of the data center flow distribution is not uniform,and the link utilization rate is not high.This paper designs a coflow routing and scheduling algorithm based on network bandwidth change,which maximizes the number of coflows that meet the deadline and minimizes the average completion time when the available bandwidth of coflow changes.