Research On Service-Oriented End-to-End QoS Routing Strategy

At present,with the rapid development of the Internet and mobile communications,many new services have been created,and network data traffic is also experiencing explosive growth.With limited bandwidth resources,how to provide Quality of Service(QoS)guarantees for various new types of services is still a problem that needs to be resolved.The traditional network architecture cannot meet the QoS requirements of different services due to functional coupling.Software Defined Network(SDN)technology is a new type of network architecture that can separate the control plane and forwarding plane of the device.The control plane can quickly collect network information,make routing decisions and send it to the forwarding plane.In recent years,the rapid development of machine learning technology has also caused extensive research on routing optimization.Using machine learning can discover the relationship between network traffic characteristics and routing strategies,allowing machine learning to replace routing tables to perform rapid routing and forwarding according to network changes.This thesis is mainly based on the SDN architecture and machine learning methods,with the core of reducing the end-to-end delay,to optimize the routing and ensure the QoS of the business.Compared with the control and forwarding coupling of traditional networks,SDN separates the control layer and the forwarding layer of network equipment,and implements the control layer in a centralized manner,which solves the shortcomings of the traditional network architecture.This thesis proposes a routing algorithm based on SDN with minimum delay constraints on the basis of existing algorithms.The algorithm uses the end-to-end average delay as the evaluation criterion for routing,collects the status information of all links in the network through SDN,matches each service flow,calculates the shortest delay path for each service flow,and satisfies the QoS of the service flow.Compared with the existing routing algorithm,this algorithm is based on the service flow for routing,which can match the OpenFlow flow table in SDN,choose the average delay as the routing criterion,and ensure that the end-to-end average delay of each service flow is the smallest and meet the QoS requirements of each service flow.In order to verify the end-to-end delay performance of the above-mentioned QoS routing algorithm,this thesis proposes an active end-to-end delay measurement method based on topology discovery.This thesis uses the Internet Control Message Protocol(ICMP)packet detection principle to find the nodes through which the data packet passes,and records the time of each node that the data packet passes through in a timestamp to calculate the delay on each link.Then this thesis constructs the network topology and routing matrix,and derives the end-to-end delay matrix based on the routing matrix and the calculated delay of each link to verify the end-to-end delay performance.Finally,this thesis designs an end-to-end QoS guarantee system based on SDN and machine learning and integrate machine learning model and delay measurement methods into the system.First,we generates the required training data through the QoS routing algorithm,and uses supervised learning to train the machine learning model.After the machine learning model is trained,it can be directly used for routing and forwarding.Then,on the basis of SDN,the link state information and service flow characteristics in the network are obtained,which are used as the input of the machine learning model to optimize routing.Finally,the end-to-end delay measurement method proposed in the thesis can verify the end-to-end delay performance.