[
https://issues.apache.org/jira/browse/BEAM-7574?focusedWorklogId=269179&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-269179
]
ASF GitHub Bot logged work on BEAM-7574:
----------------------------------------
Author: ASF GitHub Bot
Created on: 28/Jun/19 09:58
Start Date: 28/Jun/19 09:58
Worklog Time Spent: 10m
Work Description: iemejia commented on issue #8946: [BEAM-7574] fix
Combine performance for SparkRunner
URL: https://github.com/apache/beam/pull/8946#issuecomment-506678098
Run Nexmark on the current PR and saw no regressions so far and clear
improvements in Combine heavy queries. So positive about this, even if not
checked the code yet.
Command:
```
./gradlew :sdks:java:testing:nexmark:run \
-Pnexmark.runner=":beam-runners-spark" \
-Pnexmark.args="
--runner=SparkRunner
--streaming=false
--suite=SMOKE
--manageResources=false
--monitorJobs=true
--enforceEncodability=true
--enforceImmutability=true"
```
Results:
```
Master
Conf Runtime(sec) (Baseline) Events(/sec) (Baseline) Results
(Baseline)
0000 1.5 68870.5 100000
0001 0.6 172711.6 92000
0002 0.4 284900.3 351
0003 1.7 58685.4 580
0004 0.9 10893.2 40
0005 2.1 46992.5 12
0006 0.6 17006.8 103
0007 2.7 37243.9 1
0008 1.2 85324.2 6000
0009 0.5 20080.3 298
0010 0.6 161550.9 1
0011 1.0 98328.4 1919
0012 0.8 118063.8 1919
0013 0.4 267379.7 92000
0014 2.9 69710.7 184000
==========================================================================================
This PR
Conf Runtime(sec) (Baseline) Events(/sec) (Baseline) Results
(Baseline)
0000 1.3 74906.4 100000
0001 0.6 166666.7 92000
0002 0.3 392156.9 351
0003 1.6 61050.1 580
0004 0.8 12610.3 40
0005 2.1 48709.2 12
0006 0.6 17574.7 103
0007 2.4 42247.6 1
0008 1.0 98814.2 6000
0009 0.4 22421.5 298
0010 0.6 171821.3 1
0011 1.4 70028.0 1919
0012 0.8 130378.1 1919
0013 0.4 253164.6 92000
==========================================================================================
```
Note that the execution is breaking on query 14 but that seems unrelated to
the PR.
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
[email protected]
Issue Time Tracking
-------------------
Worklog Id: (was: 269179)
Time Spent: 1h 20m (was: 1h 10m)
> Spark runner: Combine.perKey performance issues
> -----------------------------------------------
>
> Key: BEAM-7574
> URL: https://issues.apache.org/jira/browse/BEAM-7574
> Project: Beam
> Issue Type: Improvement
> Components: runner-spark
> Affects Versions: 2.13.0
> Reporter: Jan Lukavský
> Assignee: Jan Lukavský
> Priority: Major
> Time Spent: 1h 20m
> Remaining Estimate: 0h
>
> Combine.perKey on current implementation uses technique of creating an
> accumulator for each input key and then merge all these accumulators
> together. Optimize this by:
> * changing accumulator from Iterable to Map, and using addInput as much as
> possible
> * try to move the window explode to pre-shuffle (add window label to key for
> non-merging windows), measure the impact, and if the impact is substantial,
> implement that for at leasit window functions assigning to single (global)
> window or single window per element (tumbling windows)
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
