Parallel filter scheduler to decrease the latency to spawn a VM in Nova
Nova Filter Scheduler: Filters are defined in nova.conf file and the request passes through this host sequentially to filter all the hosts that matches user request. So time taken is T=f1+f2+f3+.... Instead we can pass each filter through a thread and use fork-join model to pass all the hosts through different threads such that the total time taken to filter the hosts according to multi-threading concept is maximum time taken for execution of filters in threads,
Blueprint information
- Status:
- Not started
- Approver:
- None
- Priority:
- Undefined
- Drafter:
- Harshavardhan Metla
- Direction:
- Needs approval
- Assignee:
- None
- Definition:
- New
- Series goal:
- None
- Implementation:
-
Unknown
- Milestone target:
- None
- Started by
- Completed by
Related branches
Related bugs
Sprints
Whiteboard
[2020-08-27 gibi]: We discussed this on the weekly nova meeting. In summary, we don't see filter execution as a bottleneck of instance spawn. Please provide data that prove how much latency we could remove as we think now that the complexity increase is a lot bigger than the possible performance gain.