A Service Performance Aware Scheduling Approach in Containerized Cloud

Abstract

Due to the dynamic and uncertainty of users' demand for services, the resources that services depend on and the relationship between services, ensuring service performance has become a basic requirement of container cloud. There are many factors that affect service performance. Besides taking basic resources for carrying services into consideration, we also considered the delay between services caused by the relationship between services as a factor to ensure service performance, designed a container cloud dynamic monitoring framework for service performance, and proposed a service scheduling method at runtime. The design of framework can monitor service performance from two aspect of basic resources and service performance. The proposed method transforms the performance-based service scheduling problem into a planning problem that is constrained by the usage of basic resources and the delay between services. Furthermore, the proposed method generates the optimal service scheduling scheme effectively through particle swarm optimization algorithm. Compared with K8s scheduling method, the feasibility and effectiveness of this method were verified. Experimental results showed that this method could reduce the delay between services while ensuring the resource utilization and balance of the container cloud environment, so that effectively guarantee the service performance.