slightly decrease thread contention by synchronizing once rather than…

[t3hnar]

Improve Summary: slightly decrease thread contention by synchronizing once rather than "number quantiles + 1" times

[ghost]

It looks like this doesn't make a significant difference. 🙁

Benchmark:

# JMH version: 1.21
# VM version: JDK 11.0.2, OpenJDK 64-Bit Server VM, 11.0.2+9
# VM invoker: /Library/Java/JavaVirtualMachines/openjdk-11.0.2.jdk/Contents/Home/bin/java
# VM options: -javaagent:/Applications/IntelliJ IDEA CE.app/Contents/lib/idea_rt.jar=49590:/Applications/IntelliJ IDEA CE.app/Contents/bin -Dfile.encoding=UTF-8
# Warmup: 5 iterations, 10 s each
# Measurement: 4 iterations, 10 s each
# Timeout: 10 min per iteration
# Threads: 24 threads, will synchronize iterations

Please take the results with a grain of salt.

Before - 4e0e752:

Benchmark                                                           Mode  Cnt       Score        Error  Units
SummaryBenchmark.codahaleHistogramBenchmark                         avgt    4    5449.211 ±    157.085  ns/op
SummaryBenchmark.prometheusSimpleHistogramBenchmark                 avgt    4     219.256 ±     23.811  ns/op
SummaryBenchmark.prometheusSimpleHistogramChildBenchmark            avgt    4      88.585 ±      1.734  ns/op
SummaryBenchmark.prometheusSimpleHistogramNoLabelsBenchmark         avgt    4      83.989 ±      2.093  ns/op
SummaryBenchmark.prometheusSimpleSummaryBenchmark                   avgt    4     186.796 ±     15.029  ns/op
SummaryBenchmark.prometheusSimpleSummaryChildBenchmark              avgt    4      61.782 ±      5.032  ns/op
SummaryBenchmark.prometheusSimpleSummaryNoLabelsBenchmark           avgt    4      59.388 ±     12.457  ns/op
SummaryBenchmark.prometheusSimpleSummaryQuantilesBenchmark          avgt    4  530692.958 ± 627614.926  ns/op
SummaryBenchmark.prometheusSimpleSummaryQuantilesChildBenchmark     avgt    4  563807.323 ± 389413.917  ns/op
SummaryBenchmark.prometheusSimpleSummaryQuantilesNoLabelsBenchmark  avgt    4  559719.005 ± 631676.260  ns/op

After - c726b1a:

Benchmark                                                           Mode  Cnt       Score        Error  Units
SummaryBenchmark.codahaleHistogramBenchmark                         avgt    4    5651.711 ±    487.476  ns/op
SummaryBenchmark.prometheusSimpleHistogramBenchmark                 avgt    4     220.865 ±     11.554  ns/op
SummaryBenchmark.prometheusSimpleHistogramChildBenchmark            avgt    4      95.237 ±      6.059  ns/op
SummaryBenchmark.prometheusSimpleHistogramNoLabelsBenchmark         avgt    4      87.088 ±     15.551  ns/op
SummaryBenchmark.prometheusSimpleSummaryBenchmark                   avgt    4     196.351 ±     33.766  ns/op
SummaryBenchmark.prometheusSimpleSummaryChildBenchmark              avgt    4      60.853 ±      9.388  ns/op
SummaryBenchmark.prometheusSimpleSummaryNoLabelsBenchmark           avgt    4      64.209 ±      2.437  ns/op
SummaryBenchmark.prometheusSimpleSummaryQuantilesBenchmark          avgt    4  525567.096 ± 599616.355  ns/op
SummaryBenchmark.prometheusSimpleSummaryQuantilesChildBenchmark     avgt    4  724240.526 ± 864857.588  ns/op
SummaryBenchmark.prometheusSimpleSummaryQuantilesNoLabelsBenchmark  avgt    4  703944.682 ± 906458.910  ns/op

I picked the result with the smallest error from multiple benchmark runs.

Quantiles:

  .quantile(0.5, 0.05).quantile(0.9, 0.05).quantile(0.95, 0.01).quantile(0.99, 0.005)

[brian-brazil]

It likely needs a different benchmark design for this to show up.

[t3hnar]

@brian-brazil is there something which stops you from accepting this PR ? how can I improve this?

[brian-brazil]

I still have to do a proper review, however the benchmarks are not conclusive. Try a benchmark that rotates frequently.

…

On Mon 3 Jun 2019, 17:36 Yaroslav Klymko, ***@***.***> wrote: @brian-brazil <https://github.com/brian-brazil> is there something which stops you from accepting this PR ? how can I improve this? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#481?email_source=notifications&email_token=ABWJG5SCLH5D2V7NVPSGZL3PYU3AJA5CNFSM4HSDEZSKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODWZZPCI#issuecomment-498309001>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABWJG5RIOIPORX63JOE6JQTPYU3AJANCNFSM4HSDEZSA> .

[ghost]

Benchmarks with unrealistically frequent rotations by hardcoding TimeWindowQuantiles. durationBetweenRotates.

The results were consistent across many runs over the last 2 days.

Before - 4e0e752 - durationBetweenRotates=1ms:

Benchmark                                                           Mode  Cnt      Score      Error  Units
SummaryBenchmark.prometheusSimpleSummaryBenchmark                   avgt    4    180.660 ±   31.025  ns/op
SummaryBenchmark.prometheusSimpleSummaryChildBenchmark              avgt    4     59.794 ±    2.183  ns/op
SummaryBenchmark.prometheusSimpleSummaryNoLabelsBenchmark           avgt    4     62.603 ±    2.560  ns/op
SummaryBenchmark.prometheusSimpleSummaryQuantilesBenchmark          avgt    4  16668.278 ±  899.351  ns/op
SummaryBenchmark.prometheusSimpleSummaryQuantilesChildBenchmark     avgt    4  16473.560 ± 1271.724  ns/op
SummaryBenchmark.prometheusSimpleSummaryQuantilesNoLabelsBenchmark  avgt    4  16279.621 ± 1513.338  ns/op

After - c726b1a - durationBetweenRotates=1ms:

Benchmark                                                           Mode  Cnt      Score      Error  Units
SummaryBenchmark.prometheusSimpleSummaryBenchmark                   avgt    4    188.946 ±   23.181  ns/op
SummaryBenchmark.prometheusSimpleSummaryChildBenchmark              avgt    4     59.379 ±    2.877  ns/op
SummaryBenchmark.prometheusSimpleSummaryNoLabelsBenchmark           avgt    4     59.461 ±    1.214  ns/op
SummaryBenchmark.prometheusSimpleSummaryQuantilesBenchmark          avgt    4  14943.023 ± 1479.789  ns/op
SummaryBenchmark.prometheusSimpleSummaryQuantilesChildBenchmark     avgt    4   9085.545 ±  778.160  ns/op
SummaryBenchmark.prometheusSimpleSummaryQuantilesNoLabelsBenchmark  avgt    4   9181.785 ±  976.102  ns/op

Before - 4e0e752 - durationBetweenRotates=100us:

Benchmark                                                           Mode  Cnt      Score     Error  Units
SummaryBenchmark.prometheusSimpleSummaryBenchmark                   avgt    4    190.128 ±  37.522  ns/op
SummaryBenchmark.prometheusSimpleSummaryChildBenchmark              avgt    4     59.393 ±   1.586  ns/op
SummaryBenchmark.prometheusSimpleSummaryNoLabelsBenchmark           avgt    4     62.860 ±   7.727  ns/op
SummaryBenchmark.prometheusSimpleSummaryQuantilesBenchmark          avgt    4  12261.290 ± 206.572  ns/op
SummaryBenchmark.prometheusSimpleSummaryQuantilesChildBenchmark     avgt    4  12279.661 ± 117.961  ns/op
SummaryBenchmark.prometheusSimpleSummaryQuantilesNoLabelsBenchmark  avgt    4  12317.630 ± 270.580  ns/op

After - c726b1a - durationBetweenRotates=100us:

Benchmark                                                           Mode  Cnt      Score     Error  Units
SummaryBenchmark.prometheusSimpleSummaryBenchmark                   avgt    4    184.783 ±  29.702  ns/op
SummaryBenchmark.prometheusSimpleSummaryChildBenchmark              avgt    4     59.636 ±   1.406  ns/op
SummaryBenchmark.prometheusSimpleSummaryNoLabelsBenchmark           avgt    4     61.560 ±   2.485  ns/op
SummaryBenchmark.prometheusSimpleSummaryQuantilesBenchmark          avgt    4  12279.452 ±  29.655  ns/op
SummaryBenchmark.prometheusSimpleSummaryQuantilesChildBenchmark     avgt    4   7940.067 ± 590.255  ns/op
SummaryBenchmark.prometheusSimpleSummaryQuantilesNoLabelsBenchmark  avgt    4   7320.339 ± 260.343  ns/op

[brian-brazil]

Thanks!