摘要
随着云计算技术在各个领域的不断深入与发展,众多微服务之间复杂的调用关系以及相互影响也给微服务的数据监控,运行维护带来了新的挑战。目前针对多服务应用的自动伸缩方法一次只能伸缩一个瓶颈微服务,会导致瓶颈转移问题;或者依赖于服务调用图谱的静态性,不能解决动态负载下多服务应用的自动伸缩问题。本文提出了一种云计算环境下基于微服务监控系统的多服务应用自动性能分析方法,均衡地将入口服务的SLA分解为每个服务的响应时间SLO,精准获得每个服务在不超过响应时间SLO的条件下能够服务的最大流量负载和对应资源利用率,为多微服务应用集群伸缩提供了精确性能指标。
With the continuous deepening and development of cloud computing technology in various fields, the complex calling relationship and interaction between many microservices also bring new challenges to the data monitoring, operation and maintenance of microservices. The current automatic scaling method for multi-service applications can only scale one bottleneck microservice at a time, which will lead to the problem of bottleneck transfer;or relying on the static nature of the service call graph, it cannot solve the automatic scaling problem of multi-service applications under dynamic load. This paper proposes an automatic performance profiling method for multi-service applications in cloud computing environment based on a micro service monitoring system, which decomposes the SLA of the ingress service into the response time SLO of each service in a balanced manner, and accurately obtains the maximum traffic load that can be served and the corresponding resource utilization of each service under the condition that the response time SLO does not exceed SLA.
出处
《计算机科学与应用》
2022年第6期1685-1699,共15页
Computer Science and Application
