beam-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Aviem Zur (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (BEAM-1146) Decrease spark runner startup overhead
Date Sat, 25 Feb 2017 05:13:44 GMT

     [ https://issues.apache.org/jira/browse/BEAM-1146?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Aviem Zur updated BEAM-1146:
----------------------------
    Description: 
BEAM-921 introduced a lazy singleton instantiated once in each JVM (driver & executors)
which utilizes reflection to find all subclasses of Source and Coder
While this is beneficial in its own right, the change added about one minute of overhead in
spark runner startup time (which cause the first job/stage to take up to a minute).
The change is in class {{BeamSparkRunnerRegistrator}}.

  was:
BEAM-921 introduced a lazy singleton instantiated once in each JVM (driver & executors)
which utilizes reflection to find all subclasses of Source and Coder
While this is beneficial in it's own right, the change added about one minute of overhead
in spark runner startup time (which cause the first job/stage to take up to a minute).
The change is in class {{BeamSparkRunnerRegistrator}}
The reason reflection (specifically reflections library) was used here is because  there is
no current way of knowing all the source and coder classes at runtime.


> Decrease spark runner startup overhead
> --------------------------------------
>
>                 Key: BEAM-1146
>                 URL: https://issues.apache.org/jira/browse/BEAM-1146
>             Project: Beam
>          Issue Type: Improvement
>          Components: runner-spark
>            Reporter: Aviem Zur
>            Assignee: Aviem Zur
>             Fix For: 0.5.0
>
>
> BEAM-921 introduced a lazy singleton instantiated once in each JVM (driver & executors)
which utilizes reflection to find all subclasses of Source and Coder
> While this is beneficial in its own right, the change added about one minute of overhead
in spark runner startup time (which cause the first job/stage to take up to a minute).
> The change is in class {{BeamSparkRunnerRegistrator}}.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

Mime
View raw message