Clock Frequency Tiering

This is the optional clock-tiering experiment.

Use it when you want deterministic per-worker CPU lanes with per-lane max frequency caps.

warning

Do not treat this as the shared execution pool path. It is a different operating model.

Live Shape

Current intended layout:

Policy:

VM	vCPU cpuset	Emulator cpuset	Max frequency
kvm-worker-01	2-4	0-1	3000000 kHz
kvm-worker-02	5-7	0-1	2000000 kHz
kvm-worker-03	8-11	0-1	1000000 kHz

What clock-apply Does

./scripts/host-resource-management.sh clock-apply:

1. widens the live scope cpuset so libvirt pinning can take effect
2. pins each worker vCPU set to its dedicated lane
3. pins each emulator thread to 0-1
4. writes the per-CPU frequency caps under /sys/devices/system/cpu/.../cpufreq

Apply

cd /path/to/stakkr
./scripts/host-resource-management.sh clock-apply
./scripts/host-resource-management.sh clock-status

Rollback

./scripts/host-resource-management.sh clock-rollback
./scripts/host-resource-management.sh clock-status

After rollback you should see:

• virsh vcpupin back to 0-11
• virsh emulatorpin back to 0-11
• scaling_max_freq restored to cpuinfo_max_freq

tip

clock-rollback removes the clock experiment. If you want to go back to the shared-pool path afterward, run ./scripts/host-resource-management.sh shared-execution-pool-apply.

Why It Is Separate

Clock tiering does not use a shared guest execution pool. Each worker owns a dedicated lane, so it is intentionally different from the shared-pool Gold/Silver/Bronze cgroup model.