This document does not specify exact numbers, nor different type of reporting. It's only about aggregations.

The conclusion is definetly worth spreading:

"Most important conclusion in this test is that response time in ElasticSearch with aggregations is proportional to the size of the output returned, not to the number of input rows needed to be processed like in regular database. This is not just a few times faster database in all cases, but it uses different approach to data gathering, which turned out could be especially beneficial for data aggregations on big data sets."

Radek.