Research And Implementation Of A Service Expectation Based Network Routing Optimization Mechanism

Traffics belonging to different services often exist on the Internet at the same time,and traffic of different services has different characteristics,so the experiences of different services have different sensitivity to network objective indicators.For example,some services are sensitive to latency,and others are more sensitive to the packet loss rate,so there are different objects for their service optimization.How to generate the corresponding reasonable route planning based on the type of network service and meet the respective optimization expectations is the problem that this thesis focused on.Based on SDN and multi-agent deep reinforcement learning,this thesis proposes an AI based routing planning algorithm that optimizes the experiences of multi-type network services parallelly,and designs a system that uses this algorithm to plan routes and optimize multiple types of services in the network.The system can generate the forwarding path according to the service to which the traffic in the network belongs.The main work of this thesis includes two parts.Firstly,this thesis proposes a service-oriented routing planning algorithm based on SDN and multi-agent deep reinforcement learning.When the optimization target is set,the algorithm generates the forwarding path based on network status information and user service expectations for each service.Secondly,this thesis designs and implements a routing intelligent planning system based on SDN and the above algorithm,including service expectation management web page,SDN application layer module,SDN control layer module and SDN forwarding layer module.The system generates the forwarding path to optimizes the network based on the network state information collected by the lower layer module and the service expectation inputted by the user from the web page.The experimental results show that the proposed algorithm can provide corresponding routing planning strategies for the optimization goals of different services to meet the expectations of various services,which effectively improved the overall service experience quality in a network where multiple types of services coexist,and the effect is better than two comparison algorithms,the implemented system can use the algorithm to generate forwarding paths corresponding to type of service and apply it to the SDN network correctly.