Research On Flow Control Method Of Information-centric Networking Based On Multi-objective Optimization

With the continuous growth of network traffic volume and the introduction of massive applications,new requirements have been created for network architecture.Traditional TCP/IP networks are becoming increasingly difficult to meet these problems,so a new architecture of ICN(Information Centric-Networking)has emerged.ICN simplifies communication by decoupling content and location,but without host-to-host protocols,network traffic control is more complex.At the same time,the ICN network naturally supports multi-source and multipathing,and the data content of a single stream may be retrieved from multiple source-side databases or cached content,making it difficult to identify and control the status of each stream.Therefore,in ICNs,the study of network traffic control and optimization methods is an important problem faced by network researchers and network providers.In this thesis,we study the problem of ICN flow control from the following two aspects.1.In ICN,delay and throughput are the two key metrics that determine network performance.In the existing ICN traffic scheduling work,only one of the metrics,delay,and throughput,is usually considered for optimization,and even though both metrics are considered,the essence is still single-objective optimization,which leads to the situation that the other metric is only slightly improved or even decreased instead of increased during the optimization process.To address this problem,this study introduces the idea of multi-objective optimization into the traffic scheduling work and proposes the Density Sorting-based Evolution Algorithm(DSEA),which optimizes both delay and throughput metrics,and achieves a good balance between the two based on reducing delay and increasing throughput.The balance is good.Simulation experiments show that the DSEA algorithm performs best in the commonly used evaluation metrics IGD,GD,and HV compared with the other three multiobjective genetic algorithms aw GA,NSGA-II,and RVEA.2.During ICN communication,the same forwarding and routing policies are usually implemented indiscriminately for real-time and non-real-time services,and when dealing with real-time services,it is often necessary to continuously send interest packets to request data,which results in a large amount of wasted uplink bandwidth and increases the cost of communication delay.To address this problem,this study proposes a multi-node negotiation traffic control mechanism for real-time services.The near-end policy optimization algorithms(PPOs)are used to learn interactively with the environment,reward the policy when the round-trip delay decreases,and update the parameters by iterating the algorithm,so as to construct an adaptive transmission rate for real-time services and derive the probability of transmission rate selection when handling different services.Simulation experiments demonstrate that the algorithm in this study performs optimally in terms of link utilization and roundtrip delay compared with the classical congestion control algorithms AIMD,BBR,and CUBIC.