vsop-479 commented on PR #16245:
URL: https://github.com/apache/lucene/pull/16245#issuecomment-4977963641
> Maybe rerun with single segment (force merge to 1)? At least then the
results are comparable to one another, though less realistic for near-real-time
cases.
> I will add the same args to candidate_index to check.
After setting `useCMS=True` and `mergePolicy="TieredMergePolicy"` in
`example.py`, the segment counts for baseline and candidate are the same(both
23). And the performance make more sense:
# search perf
```
TaskQPS baseline StdDevQPS my_modified_version
StdDev Pct diff p-value
Respell 72.88 (3.9%) 66.79
(4.0%) -8.4% ( -15% - 0%) 0.000
Fuzzy2 106.10 (3.7%) 98.78
(5.8%) -6.9% ( -15% - 2%) 0.000
OrHighNotHigh 767.39 (9.0%) 718.70
(7.5%) -6.3% ( -20% - 11%) 0.018
AndMissingHigh 4754.32 (9.6%) 4517.80
(9.1%) -5.0% ( -21% - 15%) 0.101
Fuzzy1 90.47 (4.9%) 86.13
(5.8%) -4.8% ( -14% - 6%) 0.006
IntSet 1000.09 (11.8%) 957.58
(15.3%) -4.3% ( -28% - 25%) 0.338
HighTermTitleSort 86.95 (10.0%) 83.95
(8.7%) -3.4% ( -20% - 16%) 0.256
BrowseRandomLabelSSDVFacets 5.93 (12.2%) 5.74
(10.5%) -3.2% ( -23% - 22%) 0.384
AndHighMed 777.15 (11.1%) 754.51
(10.7%) -2.9% ( -22% - 21%) 0.409
Prefix3 1119.08 (18.0%) 1087.80
(14.8%) -2.8% ( -30% - 36%) 0.601
HighTermDayOfYearSort 291.52 (10.5%) 283.73
(9.2%) -2.7% ( -20% - 19%) 0.404
MedSpanNear 158.72 (7.3%) 155.11
(6.9%) -2.3% ( -15% - 12%) 0.325
OrNotHighLow 1841.30 (10.2%) 1800.26
(11.1%) -2.2% ( -21% - 21%) 0.519
OrHighHigh 298.07 (13.1%) 291.47
(12.8%) -2.2% ( -24% - 27%) 0.598
MedSloppyPhrase 197.58 (8.2%) 193.49
(8.4%) -2.1% ( -17% - 15%) 0.441
TermDTSort 277.63 (6.3%) 272.34
(6.0%) -1.9% ( -13% - 11%) 0.338
OrNotHighHigh 287.85 (5.9%) 282.48
(7.7%) -1.9% ( -14% - 12%) 0.403
BM25MSM2 33.42 (5.8%) 32.81
(7.1%) -1.8% ( -13% - 11%) 0.386
MedTerm 1343.32 (17.0%) 1319.14
(15.9%) -1.8% ( -29% - 37%) 0.736
BrowseDateSSDVFacets 1.28 (8.3%) 1.26
(8.1%) -1.8% ( -16% - 15%) 0.509
AndHighLow 1981.80 (10.2%) 1951.18
(9.6%) -1.5% ( -19% - 20%) 0.631
HighSpanNear 17.67 (6.0%) 17.41
(7.4%) -1.5% ( -14% - 12%) 0.500
AndHighMedDayTaxoFacets 249.07 (5.0%) 245.53
(6.7%) -1.4% ( -12% - 10%) 0.461
OrHighMed 705.24 (9.7%) 697.01
(10.1%) -1.2% ( -19% - 20%) 0.717
AndHighHigh 323.59 (11.9%) 319.94
(11.8%) -1.1% ( -22% - 25%) 0.769
LowTerm 2075.56 (19.8%) 2053.99
(17.8%) -1.0% ( -32% - 45%) 0.865
OrHighLow 1268.68 (10.3%) 1256.38
(10.3%) -1.0% ( -19% - 21%) 0.772
ConstMSM2 1922.70 (13.1%) 1906.82
(12.2%) -0.8% ( -23% - 28%) 0.841
BrowseDayOfYearSSDVFacets 7.08 (11.0%) 7.02
(11.2%) -0.8% ( -20% - 23%) 0.821
LowIntervalsOrdered 56.54 (9.6%) 56.08
(8.6%) -0.8% ( -17% - 19%) 0.783
IntNRQ 161.59 (6.0%) 160.46
(8.8%) -0.7% ( -14% - 15%) 0.775
OrNotHighMed 575.21 (8.3%) 571.27
(9.3%) -0.7% ( -16% - 18%) 0.811
HighSloppyPhrase 102.71 (9.8%) 102.01
(8.7%) -0.7% ( -17% - 19%) 0.821
LowSloppyPhrase 322.15 (6.7%) 320.17
(5.2%) -0.6% ( -11% - 12%) 0.752
LowPhrase 74.96 (5.3%) 74.58
(5.0%) -0.5% ( -10% - 10%) 0.765
Wildcard 83.22 (5.0%) 82.80
(3.7%) -0.5% ( -8% - 8%) 0.726
HighTermMonthSort 1756.28 (7.4%) 1747.85
(9.2%) -0.5% ( -15% - 17%) 0.859
OrHighNotLow 1053.05 (13.8%) 1049.42
(13.9%) -0.3% ( -24% - 31%) 0.939
AndHighHighDayTaxoFacets 23.02 (3.6%) 22.94
(4.4%) -0.3% ( -8% - 7%) 0.798
OrHighMedDayTaxoFacets 7.44 (3.2%) 7.42
(2.7%) -0.3% ( -5% - 5%) 0.759
LowSpanNear 145.76 (5.8%) 145.43
(5.9%) -0.2% ( -11% - 12%) 0.905
HighIntervalsOrdered 36.61 (8.2%) 36.61
(10.5%) -0.0% ( -17% - 20%) 0.998
MedTermDayTaxoFacets 53.66 (4.8%) 53.73
(3.6%) 0.1% ( -7% - 8%) 0.918
MedPhrase 318.83 (6.0%) 319.29
(4.9%) 0.1% ( -10% - 11%) 0.934
HighTermTitleBDVSort 48.66 (5.2%) 48.78
(5.7%) 0.3% ( -10% - 11%) 0.885
HighPhrase 111.59 (6.3%) 112.07
(8.5%) 0.4% ( -13% - 16%) 0.861
range 8167.95 (15.7%) 8205.54
(11.2%) 0.5% ( -22% - 32%) 0.917
BrowseRandomLabelTaxoFacets 4.08 (20.4%) 4.12
(16.7%) 0.9% ( -30% - 47%) 0.878
PKLookup 419.03 (3.7%) 423.69
(3.6%) 1.1% ( -5% - 8%) 0.350
OrHighNotMed 668.67 (9.7%) 679.16
(10.3%) 1.6% ( -16% - 23%) 0.630
MedIntervalsOrdered 95.92 (4.2%) 98.20
(4.0%) 2.4% ( -5% - 10%) 0.071
BrowseMonthTaxoFacets 4.99 (9.9%) 5.16
(6.2%) 3.3% ( -11% - 21%) 0.216
HighTerm 653.40 (14.7%) 680.15
(11.4%) 4.1% ( -19% - 35%) 0.338
BrowseDateTaxoFacets 4.88 (21.9%) 5.09
(20.8%) 4.3% ( -31% - 60%) 0.539
BrowseDayOfYearTaxoFacets 4.92 (22.1%) 5.14
(20.2%) 4.5% ( -30% - 60%) 0.515
BrowseMonthSSDVFacets 7.16 (15.4%) 7.49
(19.1%) 4.5% ( -25% - 46%) 0.422
```
# search gc stats
| Metric | Baseline | Candidate | Difference |
|---|---:|---:|---:|
| Elapsed time | 16,220 ms | 17,044 ms | +824 ms |
| JIT compilation time | 14,115 ms | 14,529 ms | +414 ms |
| Young GC time | 13 ms | 15 ms | +2 ms |
| Young GC collections | 9 | 9 | 0 |
| Old GC time | 20 ms | 21 ms | +1 ms |
| Old GC collections | 1 | 1 | 0 |
| Total GC time | 33 ms | 36 ms | +3 ms |
| Total GC share of elapsed time | 0.20% | 0.21% | +0.01 pp |
| Average CPU load | 637.97 / 1200 | 635.35 / 1200 | -2.62 |
# Indexing stats
| Metric | Baseline | Candidate | Difference |
|---|---:|---:|---:|
| Total docs | 33,332,620 | 33,332,620 | 0 |
| Indexing done time | 5,260,656 ms | 3,362,047 ms | -1,898,609 ms |
| Wait for merges time | 832 ms | 750 ms | -82 ms |
| Commit time | 25 ms | 23 ms | -2 ms |
| Finished time | 5,261,551 ms | 3,362,859 ms | -1,898,692 ms |
| Final segment count | 23 | 23 | 0 |
| Index size as committed | 19,486,398,890 bytes | 19,461,826,218 bytes |
-24,572,672 bytes |
| Plain text indexing throughput | 20.10 GB/hour | 31.46 GB/hour | +56.5% |
# tim store
| Metric | Baseline | Candidate | Difference |
|---|---:|---:|---:|
| `.tim` files | 23 | 23 | 0 |
| `.tim` disk usage | 1.3G | 1.2G | -0.1G |
--
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]
