Why elastic node use only 50-60 % of cpu?

[tfe2012]

Hi, I have a cluster of 1 coordinator and 16 data node. 48 shard 1 replica. 1.4 TB data.

I send a request like:

GET twitter/_search
{
  "query": {
    "term": {
      "url": "https://twitter/twit"
    }
  }
}

Kibana looks like this:

Us result I reach 4000 request/second. Is it anyway how to improve this number?

[warkolm (Mark Walkom)]

Improve it to what? Is the search slow?
Why do you want 100% of your CPU being used?

[tfe2012]

I want reach 20 000 - 40 000 request's per second.

But don't know how. I need

1. change config ?
2. add more data nodes ?
3. add more coordinators ?

[warkolm (Mark Walkom)]

How are you send the requests to Elasticsearch?
What sort of data is it?
Is that query all you are doing?
Have you done benchmarking?

[tfe2012]

1. sending using java client . port 9300

2. Documents are like:

{
        "_index" : "twitter",
        "_type" : "twitter",
        "_id" : "5280F4F01D6A6F7CE386E1D135CD3162",
        "_score" : 15.434762,
        "_source" : {
          "urls" : [
            "https://twitter/twit",
            "/twit"
          ],
          "fullname" : "Test name",
          "url" : "https://twitter/twit"
        }
      }

3. I scroll index and only send tem request.

4. no