[
https://issues.apache.org/jira/browse/TAJO-1130?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14367492#comment-14367492
]
ASF GitHub Bot commented on TAJO-1130:
--------------------------------------
Github user jihoonson commented on a diff in the pull request:
https://github.com/apache/tajo/pull/423#discussion_r26684907
--- Diff: tajo-common/src/main/java/org/apache/tajo/conf/TajoConf.java ---
@@ -315,6 +315,9 @@ public static int setDateOrder(int dateOrder) {
$GROUPBY_MULTI_LEVEL_ENABLED("tajo.dist-query.groupby.multi-level-aggr", true),
+ // WARN "tajo.yarn-rm.parallel-task-runner-launcher-num" should be set
enough to avoid deadlock
+ $QUERY_EXECUTE_PARALLEL_MAX("tajo.query.execute.parallel.max", 2),
--- End diff --
There are a lot of remaining issues to fully support concurrent stage
execution. One of the representative issues is resource scheduling. However, as
you already know, our new resource scheduler which can schedule resources for
multiple queries and stages is on-going work. So, we cannot handle this issue
now.
Thus, even though this feature is still nice, it should be disabled by
default. In other words, the default value of parallelism should be 1.
Additionally, this is related to the concurrent execution of multiple
stages. So, it would be better if the name is intuitive. How about
```tajo.querymaster.stage.parallel-execution.max-num```?
> Concurrent execution of independent execution blocks
> -----------------------------------------------------
>
> Key: TAJO-1130
> URL: https://issues.apache.org/jira/browse/TAJO-1130
> Project: Tajo
> Issue Type: Improvement
> Components: query master
> Reporter: Jihoon Son
> Assignee: Navis
>
> Currently, Tajo can execute ExecutionBlocks one by one even though there
> remain enough resources to execute two or more ExecutionBlocks.
> We can improve the query processing performance by executing two or more
> independent ExecutionBlocks if possible.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
