The Placement API service was introduced in Nova in Mitaka and in Ocata is now a required part of an OpenStack deployment. This presentation will cover the design of the resource provider data modeling that acts as the foundation for new scheduling and resource tracking functionality designed to serve Nova and other OpenStack service projects.
Attendees will walk through the different resource tracking subsystems in Nova, learning how information is passed between Nova compute agents, the Placement API, the Nova scheduler and conductor, and the hypervisor systems. Understand how providers of shared resources, such as shared storage pools, work with the system to reflect accurate resource reporting and scheduling decision-making.
Attendees will learn about the following:
- How and what information is passed between Nova's many internal subsystems that track resource inventory and usage
- The implementation of Placement API service; how it's different from the Nova API service and how it's similar
- The data modeling schema behind the Placement service
- Crazy Acts of SQL Science: see the SQL Join-fu behind the Placement service's decision-making queries
- The way that accounting of resources is made in a transactional manner that tolerates multiple concurrent decision-makers
