The problem

What does a scheduler optimize when its users have contradictory requirements?

What current research shows

The Linux scheduler exposes multiple classes, policies, cgroups, affinity controls, utilization clamps, and extensibility precisely because workload requirements differ. A general default must balance responsiveness, throughput, fairness, energy, and broad compatibility.

Where the evidence stops

The existence of a compromise doesn’t prove a replacement is better. A specialized policy must disclose its workload and failure modes.

What Valen Systems is testing

Define workload classes, measure the tradeoffs separately, and make policy selection explicit instead of hiding specialization inside a universal claim.

Sources

Linux scheduler: EEVDFLinux scheduler: utilization clampingLinux extensible scheduler class