实时应用中微服务扩展策略的优化OA

Dynamic load balancing drives the development of elastic scaling techniques.However,real-time applications under microservice architectures still suffer from hysteresis when using the native scaling strategy of Kubernetes,and a single scaling metric also has limitations.Therefore,we optimized the HPA strategy,and deployed a dynamic scaling window to cope with load perturbation.We proposed an adaptive scaling strategy to satisfy multiple scaling metrics.Meanwhile microservices proactive scaling was realized to meet the high timeliness demand of the response results through ARIMA model within the scaling window.The channel of MAPE was designed to implement the adaptive scaling strategy.Experimental results show that this strategy eliminates the ping-pong effect of the number of pods and reduces the average response time by 21.3%compared with the native HPA strategy,which is more suitable for latency-sensitive services.