The problem

Which idle CPU should receive a task when the CPUs don’t have the same capacity or energy cost?

What current research shows

Linux's capacity-aware scheduling model represents CPUs with different capacities and allows utilization clamps to bias placement and frequency. That makes the tradeoff explicit: a small task may fit on an efficiency core, while a latency-sensitive task may need a faster CPU even when its historical utilization is low.

Where the evidence stops

Hardware topology and frequency behavior vary. A placement policy must be evaluated on the target machine instead of generalized from a single heterogeneous system.

What Valen Systems is testing

Evaluate declared latency needs against capacity, frequency transition time, energy, and thermal behavior across heterogeneous machines.

Sources

Linux scheduler: capacity-aware schedulingLinux scheduler: utilization clampingLinux scheduler: Energy Aware Scheduling