[
https://issues.apache.org/jira/browse/HIVE-13226?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15184667#comment-15184667
]
Prasanth Jayachandran commented on HIVE-13226:
----------------------------------------------
The problem is from user perspective, printing methods is not really helpful.
"Analyze" for example have no context. Its also combination of semantic
analyze, logical optimization and task compilation. Also it misses some steps
in-between which will be useful to find where time is spent. For example, time
between TezBuildDag and TezSubmitToRunningDag is not accounted which is the
time taken for resource localization, session restart etc.
"DAG Submit to DAG Accept" -> "DAG Submit to Accept".. is that any better?
> Improve tez print summary to print query execution breakdown
> ------------------------------------------------------------
>
> Key: HIVE-13226
> URL: https://issues.apache.org/jira/browse/HIVE-13226
> Project: Hive
> Issue Type: Improvement
> Affects Versions: 2.1.0
> Reporter: Prasanth Jayachandran
> Assignee: Prasanth Jayachandran
> Attachments: HIVE-13226.1.patch, sampleoutput.png
>
>
> When tez print summary is enabled, methods summary is printed which are
> difficult to correlate with the actual execution time. We can improve that to
> print the execution times in the sequence of operations that happens behind
> the scenes.
> Instead of printing the methods name it will be useful to print something
> like below
> 1) Query Compilation time
> 2) Query Submit to DAG Submit time
> 3) DAG Submit to DAG Accept time
> 4) DAG Accept to DAG Start time
> 5) DAG Start to DAG End time
> With this it will be easier to find out where the actual time is spent.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
