Using cgroups/cpusets to save power
Alternative to sched_mc and hotplug
Blueprint information
- Status:
- Complete
- Approver:
- Daniel Lezcano
- Priority:
- High
- Drafter:
- Daniel Lezcano
- Direction:
- Approved
- Assignee:
- Daniel Lezcano
- Definition:
- Approved
- Series goal:
- None
- Implementation:
- Implemented
- Milestone target:
- None
- Started by
- Daniel Lezcano
- Completed by
- Daniel Lezcano
Whiteboard
Discussion results:
* identify all interrupts waking up a cpu
=> https:/
* reduce the number of interrupt
* timer irq affinity
=> https:/
* timer and workqueue migration to non idle cpus
=> https:/
* solution is suitable when:
* cpuset groups cpus per cluster
* cpuidle supports deep idle state (powerdown)
* solution is suitable for b.L to:
* move the policy in the userspace
* low latency to switch the cluster is not critical
* a prototype has been made
* a study has been done:
=> elinux.