How to simulate small frequent updates parallel to constant queries

Hi,

when I am benchmarking my querying with Rally, I would like to simulate also the fact that I get some frequent but small updates. Not sure what would be the best way to achieve this?

Say I want to have a throughput of 100 queries with a sparkling of some very small (handful of docs) updates every 20 to 30 seconds (with some randomization if possible)

How would you go about it? For now I am thinking of something like:

{
    "name": "search_all",
    "default": true,
    "schedule": [
       //delete, bulk index etc...
  
      {
          "parallel": {
            "warmup-iterations": 10,
            "iterations": 1000,
            "completed-by": "delta-bulk",
            "tasks": [
                {
                  "name": "delta-bulk",
                  "operation": {
                      "operation-type": "bulk",
                      "bulk-size": 2,
                      "target-throughput": 0.1
                  },
                  "clients": 1
                },
                {
                    "operation": "randomized-query",
                    "clients": 4,
                    "iterations": 1000,
                    "target-throughput": 1
                }
            ]
          }
        }

    ]
}

but my guess this can be improved? Not sure if I should be using conflicts (and how) etc

thanks

Hello,

You can use the conflicts property of the bulk operation.

You can control the frequency of updates with conflict-probability in propotion to the target-throughput you've set for bulk. Additional properties that you can leverage are for what to do on conflict (index or update) via on-conflict and, finally, whether conflicts should be biased towards more recent or older ids (see recency).