China Mobile's public cloud is building on OpenStack. We start the work in early 2015, until now, we have totally build three regions, nearly more than 100,000 instances running on it. In the growing of business, We have faced with many challenges in architecture, high concurrency and super-large scale region. This topic will show the best practices in building and running public cloud on OpenStack in China Mobile.
1. How to handle the hundreds of concurrency on user-defined images(may not have volume cache) in the strict time limit?
2. When the resource of the single OpenStack cluster has been sold out, how to do capacity expansion smoothly, and how many instances are suitable for a nova-compute or cluster?
3.How to design the relations between ceph clusters and OpenStack clusters?
4. Super-large scale regions, does cellv2 or cascading OpenStack fit our needs?
In this topic, the attendees will learn:
1. Architectural evolution in different periods of public cloud that based on OpenStack.
2. Best practices in the continuous increase of public cloud scale, how to build super-large scale region.
3. High concurrency on hot resources, the best way to do preheat.