[
https://issues.apache.org/jira/browse/ASTERIXDB-3611?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18003555#comment-18003555
]
ASF subversion and git services commented on ASTERIXDB-3611:
------------------------------------------------------------
Commit d3b76c742127e678c18e18a715f88192bd3d02b5 in asterixdb's branch
refs/heads/master from Ian Maxon
[ https://gitbox.apache.org/repos/asf?p=asterixdb.git;h=d3b76c7421 ]
[ASTERIXDB-3611][RT] Add more stats to query profile
- user model changes: yes
- storage format changes: no
- interface changes: yes
Details:
Add frame counts and average tuples per frame to the
query profile. Also add the raw size of all tuples
forwarded from the operator.
Additionally, improve the calculation of average tuple
size to be a normal average rather than a running average.
This should be fine, because the number of bytes flowing
through and operator should be less than ~8 EB.
Furthermore, pull up all of these stats, along with the
tuple size stats, to the summarized profile in the query
plan. The averages represented in the summarized plan are
a weighted average of each operator's average, proportional
to the cardinality of that instance versus the total across
all partitions.
Ext-ref: MB-66797
Change-Id: I524cbfb1685ad3c7ed269c2db48482e5aba1ce61
Reviewed-on: https://asterix-gerrit.ics.uci.edu/c/asterixdb/+/19809
Integration-Tests: Jenkins <[email protected]>
Tested-by: Jenkins <[email protected]>
Reviewed-by: Ali Alsuliman <[email protected]>
> Add frame counts and average tuple counts per frame to runtime profile
> ----------------------------------------------------------------------
>
> Key: ASTERIXDB-3611
> URL: https://issues.apache.org/jira/browse/ASTERIXDB-3611
> Project: Apache AsterixDB
> Issue Type: Improvement
> Components: API - HTTP API, RT - Runtime
> Reporter: Ian Maxon
> Assignee: Ian Maxon
> Priority: Major
> Labels: triaged
>
> These additional profiling stats are very useful in identifying situations
> where large tuples are causing frames to be sent under-full, and therefore
> causing the query to run nearly iteratively
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
