Why is no warmup defined in event-data-track challenge elasticlogs-querying

[asp]

Hi @Christian_Dahlqvist / @danielmitterdorfer,

I made quite some tests with the event-data-track.

My testrun was like the following:

• restart elasticsearch
• import 250 mio events via challenge elasticlogs-1bn-load
• do this 3x
  • restart elasticsearch
  • run challenge elasticlogs-querying.
• do this 5x
  • restart elasticsearch
  • run challenge combined-indexing-and-querying
• do this 2x
  • restart elasticsearch
  • run challenge [elasticlogs-querying](https://github.com/elastic/rally-eventdata-

When I check the service time of kibana queries during elasticlogs-querying I can see a massive spike at the beginning.

The image shows the service time of the 1st three iterations of elasticlogs-querying.

For me it looks like some warmup issue which occurs after massive writing. (maybe disk cache by OS?).

I just want to ask, why the warmup period is explicitly set to 0.

   {
      "parallel": {
        "warmup-time-period": 0,
        "time-period": 1800,
        "clients": 4,
        "tasks": [
          {
            "operation": "relative-kibana-content_issues-dashboard_50%",
            "target-interval": 30
          },
          {
            "operation": "relative-kibana-content_issues-dashboard_75%",
            "target-interval": 90
          },
          {
            "operation": "relative-kibana-traffic-dashboard_50%",
            "target-interval": 40
          },
          {
            "operation": "relative-kibana-discover_50%",
            "target-interval": 60
          }
        ]
      } 

Is there a deeper meaning behind that? What is it? I just would like to understand the decision

Regards, Andreas

[Christian_Dahlqvist]

If I remember correctly I explicitly set it that way as I wanted to see all data points. As you correctly point out having a short warmup might have been a good idea.