Research On Dynamic Traffic Control Mechanism For Cloud Services

With the popularity and rapid development of the Internet,cloud computing services have played a very important role in the development of the Internet.At the same time,cloud computing services have brought about the diversification and complexity of network traffic.By using an effective traffic control mechanism in the network,network quality of service(QoS)and network delay can be guaranteed to provide a better user experience.However,in cloud service systems,traditional traffic control technologies lack real-time dynamics and hierarchy.Therefore,the resear-ch focus of this thesis is to design a dynamic multi-level traffic control mechanism and optimization of the traffic control mechanism.The main research contents are as follows:Firstly,this thesis designs a cloud service monitoring platform based on the requirements of cloud service system,which is used to collect the service node data.Based on this,a dynamic multi-level traffic control mechanism is designed.Firstly,in the service cluster,the multi-dimensional dynamic feedback load balancing strategy is adopted to report the comprehensive load of the service node to the cloud service monitoring platform in real time.When the network traffic enters the cluster,the service node with the smallest integrated load is selected for traffic distribution;Within the service node,the dynamic lifting level traffic limiting mechanism is adopted to dynamically control the network traffic through the health of the service node and the dynamic traffic limiting threshold.Tests in the cloud service system show that the dynamic two-level traffic control mechanism can effectively improve the resource utilization of the service node and the RPS(Requests Per Second)of the service node.Secondly,this thesis establishes a traffic allocation model based on M/M/1 according to the request allocation and processing scenarios of the cloud service system.According to the integrated load of each service node,a traffic allocation algorithm based on the expected arrival rate is proposed.Through the test in the actual cloud service system and the theoretical derivation in the queuing model,the relationship between the expected arrival rate and the integrated load is obtained.Based on this relationship,the expected number of requests of the service node is calculated according to the integrated load,and the network traffic is reasonably allocated in the service cluster.According to the test in the actual cloud service system,the optimization scheme of the flow control mechanism based on queuing theory can improve the resource utilization rate and request processing efficiency of the service node.