Research and Improvement of Dynamic Highly Available Load Balancing Algorithm for Microservices

In microservices, in order to maintain the efficiency and reliability of the system under high concurrency and to solve the problem that the load balancer polling method in microservices cannot adapt to the high concurrency and high load situation, an algorithm DWLOAD (Dynamic Weight Loading Algorithm) is proposed for the dynamic weight server based on virtual nodes in conjunction with the actual application scenario. The algorithm introduces the concepts of virtual nodes and real-time weights based on a smooth polling load algorithm. The DWLOAD algorithm calculates the number of recent requests, response time, and throughput of the same microservice module deployed in different servers to obtain the real-time weights of each server and records them in the weights table, and uses the smooth polling loading algorithm to initialize virtual nodes in batches. The virtual nodes are initialized in batches using a smooth polling load algorithm. Through simulation experiments, the QPS processing capability of DWLOAD is improved by about 50% compared with the smooth polling algorithm in large-scale scenarios, which effectively improves the efficiency and reliability of load balancing in high-load situations in microservices and reduces the comprehensive response time of the system.