Nomad scheduler tick interval limits task dispatch latency

Goal

Determine whether the current scheduler tick interval is a bottleneck for task dispatch latency before applying optimizations such as resource re‑allocation or job resubmission.

Constraints & Uncertainty

Nomad’s best‑fit decreasing bin‑packing algorithm can unevenly distribute resources, especially in heterogeneous workloads. The default scheduler tick interval of 100 ms directly influences how quickly the scheduler reacts to resource changes, but decreasing it increases CPU overhead. There is no built‑in preemption; freeing resources requires manual intervention or job resubmission.

Specific Questions

• What is the measured dispatch latency for a benchmark job under the current 100 ms tick interval?
• How does reducing the tick interval affect overall CPU utilization on the Nomad servers?
• Is there a documented or recommended range for the scheduler tick interval that balances responsiveness and overhead?