[
https://issues.apache.org/jira/browse/DRILL-5187?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15960358#comment-15960358
]
Khurram Faraaz commented on DRILL-5187:
---------------------------------------
[~zfong] TPCDS query 72 takes 24 mins on latest apache Drill master 1.11.0 on
three node cluster, can some one please look into this. On Drill 1.10.0 same
query took 32 mins, so there is a slight improvement, but we may want to
investigate and improve the execution time of tpcds query 72.
apache Drill 1.11.0
git.commit.id.abbrev=06e1522
TPC-DS query 72 profile: ( on SF1 parquet data)
TOTAL FRAGMENTS: 11
DURATION: 24 min 2.816 sec
PLANNING: 1.567 sec
QUEUED: Not Available
EXECUTION: 24 min 1.249 sec
> TPC-DS query 72 takes 32 minutes to complete
> --------------------------------------------
>
> Key: DRILL-5187
> URL: https://issues.apache.org/jira/browse/DRILL-5187
> Project: Apache Drill
> Issue Type: Bug
> Components: Execution - Flow
> Affects Versions: 1.10.0
> Environment: 4 node cluster
> Reporter: Khurram Faraaz
> Priority: Critical
> Attachments: quert72_query_plan.txt, query72_drillbit.log,
> query72_JSON_profile.txt, tpcds_query72_profile.png
>
>
> TPC-DS query takes close to 32 minutes to complete execution on a 4 node
> CentOS cluster, on Drill 1.10.0 against SF1 data.
> Attached here are,
> - query plan for query 72
> - drillbit.log (has only query 72 information)
> - JSON profile for query 72
> - screen shot of fragment information from profiles tab on UI
>
> HASH_JOIN operator takes 21minutes (fragment : 04-xx-53).
> HASH_JOIN operator seems to be taking longer, it can be verified by looking
> at the profile information.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
