According to Elasticsearch benchmarks, ES 7.15.0 became slower than ES 6.8.14 in term, pharse, mothly_agg queris in pmc track, but faster in most of queries (except hourly_agg) in http logs track.
So, which changes of Elasticsearch or lucene make the degrade in full text search and upgrade in http-logs?
https://elasticsearch-benchmarks.elastic.co/#tracks/pmc/release
https://elasticsearch-benchmarks.elastic.co/#tracks/http-logs/release
PS:
It makes no sense that es7 use less heap memory (index files off heap), but take a longer gc time.
es 7.15.0:
| Total Young Gen GC time | | 2.31 | s |
| Total Young Gen GC count | | 73 | |
| Total Old Gen GC time | | 0 | s |
| Total Old Gen GC count | | 0 | |
| Store size | | 19.1106 | GB |
| Translog size | | 3.07336e-07 | GB |
| Heap used for segments | | 0.417774 | MB |
| Heap used for doc values | | 0.0691032 | MB |
| Heap used for terms | | 0.238342 | MB |
| Heap used for norms | | 0.0405884 | MB |
| Heap used for points | | 0 | MB |
| Heap used for stored fields | | 0.0697403 | MB |
es6.8.20:
| Total Young Gen GC time | | 0.936 | s |
| Total Young Gen GC count | | 59 | |
| Total Old Gen GC time | | 0 | s |
| Total Old Gen GC count | | 0 | |
| Store size | | 20.3933 | GB |
| Translog size | | 2.67861 | GB |
| Heap used for segments | | 20.3588 | MB |
| Heap used for doc values | | 0.114174 | MB |
| Heap used for terms | | 18.3177 | MB |
| Heap used for norms | | 0.0384521 | MB |
| Heap used for points | | 0.00925446 | MB |
| Heap used for stored fields | | 1.87921 | MB |