/sig scheduling

When running scheduler-perf on my PC, it fails to record SchdulingThroughput in some scenarios. The specs of My PC: 12th Gen Intel(R) Core(TM) i5-1240P 1.70 GHz 16threads, 24GB RAM.

The dataItem is like this:

    {
      "data": null,
      "unit": "",
      "labels": {
        "Metric": "not applicable",
        "Name": "BenchmarkPerfScheduling/SchedulingWithNodeInclusionPolicy/500Nodes/namespace-2",
        "event": "not applicable",
        "extension_point": "not applicable",
        "plugin": "not applicable",
        "result": "not applicable"
      }
    },

throughputCollector measures the throughput at 1-second intervals. The issue occurs when all of pods are scheduled before 1 second passes.

We would have several solutions, but they have each pros and cons.

• Just reducing the interval time
  • pros: Simple
  • cons: It could affect the scheduler performance
• Calcurate the overall throughput based on the number of all pods and elapsed time
  • pros: No side effects
  • cons: Difficult to accurately calcurate

/cc @sanposhiho @macsko