Research And Implementation Of Micro-service Load Balancing Based On Dynamic Weights

With the rapid development of Internet technology,micro-service has become a popular software architecture.The microservice architecture can split a large system into multiple small microservices,each of which is responsible for specific functions.In order to obtain better service effect,microservices can be deployed to different servers to form microservices cluster.Under the cluster architecture,the current load balancing algorithm is easy to cause the server resource allocation imbalance,the user request can not be answered in time and so on,so that the effect of the cluster can not be brought into full play.This paper focuses on the research of load balancing algorithm under micro-service cluster.Through dynamic acquisition and weight distribution of server resources,the existing load balancing algorithm is optimized.The main work is as follows:First,after the client sends the request,the service calls the overall process in the micro service cluster,and the process is divided into the client,processing area and micro service cluster for processing.The dynamic weight load balancing algorithm studied in this paper first saves the acquired data,then predicts the resource data of the next second using the XGBoost model,and finally calculates the load value of the server to select the optimal server according to the load value.Add the resource acquisition module to the server,transfer the resource data to the processing area,balance the response to the client and the load,realize the service call in the cluster,improve the resource utilization,increase the execution efficiency and reduce the service response time under high concurrency.Secondly,the simulation experiments are conducted.Select Spring Cloud series to build this micro service cluster,select Eureka to establish a registration center,and build a gateway with Gateway.After cluster-trace-v2018 of Ali dataset,we obtain the optimal parameter value,compare the optimized XGBoost algorithm with the random forest,prove that the optimized prediction model works well,and realizes the present algorithm.Finally,combined with the dynamic weight load balancing algorithm,the instrument management system based on microservices is designed,which realizes the functions of instrument management,instrument scrapping management,instrument maintenance management,responsible person management,instrument allocation management,instrument lease management,asset transfer management,and the instrument management system is deployed to the cluster for experiments.the results show that the dynamic weight load balancing algorithm in this paper has good applicability and certain application value.Compared with common polling algorithm and weighted polling algorithm,it is found that the dynamic weight load balancing algorithm studied in this paper is the best in high concurrent requests,which reduces the response time of API and improves the service ability of micro service cluster.