EEVDF scheduling introduces latency spikes for interactive workloads

EEVDF scheduling introduces latency spikes for interactive workloads

Goal: Evaluate whether the switch from CFS to EEVDF in recent kernels causes measurable latency increases for latency‑sensitive applications such as terminal multiplexers or text editors.

Constraints: Kernel 6.6+ uses EEVDF by default; older kernels still use CFS. Workload consists of short, blocking I/O operations interleaved with CPU bursts. Uncertainty: how the scheduler’s deadline calculation interacts with task priority and weight in this scenario.

Questions:

• What kernel configuration or command‑line option toggles EEVDF vs CFS?
• How does EEVDF compute a task’s deadline relative to its weight and virtual runtime?
• What observable metrics (e.g., /proc/sched_debug, perf sched) can confirm whether latency spikes are due to EEVDF behavior?