Added support for avro-util's fastserde. #85
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
FelixGV
force-pushed
the
fast_avro_support
branch
from
a4c4f7a to
2afdc87
Compare
FelixGV
commented
Jun 17, 2020
Comment on lines
56
to
67
| System.err.println("-----------------------------------------------------------------------------"); | ||
| System.err.println(serializer.getName()); | ||
| System.err.println("min:" + measurementVals[0]); | ||
| System.err.println("1/4:"+measurementVals[count/4]); | ||
| System.err.println("1/2:"+measurementVals[count/2]); | ||
| System.err.println("3/4:"+measurementVals[count/4*3]); | ||
| System.err.println("max:"+measurementVals[count-1]); | ||
| System.err.println("average:"+ avg +"ns deviation:"+(avg-measurementVals[count/2])+"ns, total iteration:" + count * iterations); | ||
| System.err.println(testCaseName + ": " + serializer.getName()); | ||
| System.err.println("min: " + printLatency(measurementVals[0])); | ||
| System.err.println(printNthPercentileLatency(count, 25)); | ||
| System.err.println(printNthPercentileLatency(count, 50)); | ||
| System.err.println(printNthPercentileLatency(count, 75)); | ||
| System.err.println(printNthPercentileLatency(count, 99)); | ||
| System.err.println("max: " + printLatency(measurementVals[count - 1])); | ||
| System.err.println("avg: " + printLatency(avg)); | ||
| System.err.println("deviation: " + printLatency(avg-measurementVals[count/2])); | ||
| System.err.println("total iterations: " + count * iterations); | ||
| System.err.println("-----------------------------------------------------------------------------"); |
There was a problem hiding this comment.
In case anyone is wondering, the new output looks like this:
test-time 3333 iteration 2000
-----------------------------------------------------------------------------
Create: flatbuffers
min: 47.2 ns
p25: 50.1 ns
p50: 51.4 ns
p75: 55.8 ns
p99: 100.5 ns
max: 1640 ns
avg: 58.4 ns
deviation: 6.9 ns
total iterations: 56930000
-----------------------------------------------------------------------------
test-time 10000 iteration 2000
-----------------------------------------------------------------------------
Serialize: flatbuffers
min: 1430.5 ns
p25: 1548.7 ns
p50: 1595.8 ns
p75: 1655.8 ns
p99: 2228.5 ns
max: 3380.5 ns
avg: 1634.1 ns
deviation: 35.7 ns
total iterations: 6118000
-----------------------------------------------------------------------------
test-time 10000 iteration 2000
-----------------------------------------------------------------------------
Deserialize: flatbuffers
min: 923.9 ns
p25: 998.4 ns
p50: 1027.9 ns
p75: 1074 ns
p99: 1624.9 ns
max: 11662.7 ns
avg: 1067.2 ns
deviation: 38.1 ns
total iterations: 9358000
-----------------------------------------------------------------------------
This is a purely cosmetic change so I'm happy to drop it from the PR, if you think it's inappropriate. I am mostly interested in merging the enhanced Avro test coverage.
FelixGV
force-pushed
the
fast_avro_support
branch
from
2afdc87 to
737f57d
Compare
Added a "manual" optimization mode for Avro where object reuse is exercised. The existing avro test cases are slightly refactored but unchanged from the perspective of performance characteristics. Fixed some of the SerFeatures characteristics for Avro. Also tweaked the TestCaseRunner output format so that it looks a bit more polished: added a time unit (ns) to all latency measures, made each metric have its own line, limited latency to one digit (i.e. 100 pico granularity), added the test scenario (create, ser, deser) next to the serde implementation name. Finally, added a p99 latency metric to the output.
FelixGV
force-pushed
the
fast_avro_support
branch
from
737f57d to
806aa59
Compare
|
Thank you Felix. 🙂 The optimisations and adjustments make sense. |
|
Wow thanks, that was fast :D ! How does the results page update work? Do you have a given hardware that you run this on and update yourself? Or do you expect the community to submit a PR to update the results page? |
|
You put in effort, and thank me for accepting the fruits. 😄 I hate how people get stuck on GitHub. The owner kindly allowed me to speed things up. So yes, more community work please. I've opened issue #82 because my hardware is too old. The results page is a public wiki. |
|
Thanks, I add a comment on the other issue. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Added a "manual" optimization mode for Avro where object reuse is
exercised. The existing avro test cases are slightly refactored
but unchanged from the perspective of performance characteristics.
Fixed some of the SerFeatures characteristics for Avro.
Also tweaked the TestCaseRunner output format so that it looks a bit
more polished: added a time unit (ns) to all latency measures, made
each metric have its own line, limited latency to one digit (i.e.
100 pico granularity), added the test scenario (create, ser, deser)
next to the serde implementation name. Also added a p99 latency
metric.