I have an ES cluster with a configuration of 32c128g and two machines. One fragment, one master and one slave. There are a total of 600 data points in the ES index, estimated to be within 5M (each data point is not large). Why can’t the query performance reach 20000 qps per machine under such a configuration? Is there a problem with using it somewhere?

I have an ES cluster with a configuration of 32c128g and two machines. One fragment, one master and one slave. There are a total of 600 data points in the ES index, estimated to be within 5M (each data point is not large). Why can’t the query performance reach 20000 qps per machine under such a configuration? Is there a problem with using it somewhere?