navneet1v commented on PR #16815:
URL: https://github.com/apache/lucene/pull/16815#issuecomment-6126566591
Here are the benchmarks results of the heap comparison:
Dataset used: 1M, 1024D
```
PARAMS = {
"ndoc": (1_000_000,),
"maxConn": (16,),
"beamWidthIndex": (256,),
"fanout": (100,),
"numMergeWorker": (8,),
"numMergeThread": (4,),
"numSearchThread": (4,),
"encoding": ("float32",),
"metric": ("dot_product",),
"quantizeBits": (32,),
"topK": (100,),
"forceMerge": (True,),
"nquery": (1000,),
"reindex": (True,),
"offHeapGraphBuild": (True,False),
}
```
```
Force merge to 1 segment (1M docs, 1 GB heap):
┌─────────────────────────────────────┬──────────────────────┬─────────────────────┐
│ │ Candidate (off-heap) │
Baseline │
├─────────────────────────────────────┼──────────────────────┼─────────────────────┤
│ Live heap after GC, p50 / p90 / max │ 91 / 91 / 91 MB │ 375 / 998 /
1024 MB │
├─────────────────────────────────────┼──────────────────────┼─────────────────────┤
│ Full GCs │ 0 │ 2
│
├─────────────────────────────────────┼──────────────────────┼─────────────────────┤
│ Concurrent mark cycles │ 0 │ 6
│
├─────────────────────────────────────┼──────────────────────┼─────────────────────┤
│ GCs per minute │ 3.0 │ 7.2
│
├─────────────────────────────────────┼──────────────────────┼─────────────────────┤
│ Total GC pause / longest pause │ 0.02 s / 1.7 ms │ 0.52 s /
81.7 ms │
└─────────────────────────────────────┴──────────────────────┴─────────────────────┘
```
```
Speed and quality were unchanged:
┌───────────────────┬───────────┬───────────┐
│ │ Candidate │ Baseline │
├───────────────────┼───────────┼───────────┤
│ Indexing │ 374 s │ 373 s │
├───────────────────┼───────────┼───────────┤
│ Background merges │ 405 s │ 419 s │
├───────────────────┼───────────┼───────────┤
│ Force merge │ 321 s │ 328 s │
├───────────────────┼───────────┼───────────┤
│ Recall │ 0.951 │ 0.952 │
├───────────────────┼───────────┼───────────┤
│ Search latency │ 2.41 ms │ 2.43 ms │
├───────────────────┼───────────┼───────────┤
│ Index size │ 3985.6 MB │ 3985.5 MB │
└───────────────────┴───────────┴───────────┘
```
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
