Efficient resource pool management is crucial to achieve high scalability and elasticity in a cloud environment. In this presentation, users from China Mobile will share their experiences on managing thousand of VMs using Senlin service. Topics like integrating Ceilometer/Gnocchi with Senlin to support auto-scaling, optimizing OpenStack services to achieve high concurrency, and deploying Senlin on multiple hosts to support service HA will be discussed.
We will first introduce our OpenStack environment and some typical use cases. Then we will analyze performance bottlenecks of cluster provisioning and scaling and how to solve them by optimizing OpenStack and Senlin service. Performance data and benchmark result will be given to verify our optimization. And we will also talk about service HA of Senlin applied in our system. Attendees will learn some experiences on managing large scale VM clusters and the way to probe performance issues which could happen at different layers and how to address them by optimizing the service design and configuration.