[
https://issues.apache.org/jira/browse/HIVE-549?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12769433#action_12769433
]
He Yongqiang commented on HIVE-549:
-----------------------------------
Agree on adding a new TaskRunner which extends Thread.
I think we can define several states for each TaskRunner.
An simple example can be:
1. iterate the task queue and for each task lunch a new TaskRunner for it
2. For a TaskRunner serving for a special task, first check the states of all
of his parents
2.1 If any of his parents got dead(Failed), Fail itself
2.2 if all parents are successful, lunch itself and mark itself as running
2.3 if any of its parents are not running or are not finished/failed, mark
itself as pending. And also add itself in a waiting queue of its not finished
parent.
3. If a task successfully finished or failed, notify all children in its
waiting queue of whethe it successed or failed.
Obviously, this simple algorithm has one problem: each task got one dedicated
thread allocated for it. It will be better if we can compress thread numbers.
> Parallel Execution Mechanism
> ----------------------------
>
> Key: HIVE-549
> URL: https://issues.apache.org/jira/browse/HIVE-549
> Project: Hadoop Hive
> Issue Type: Wish
> Components: Query Processor
> Reporter: Adam Kramer
> Assignee: Chaitanya Mishra
>
> In a massively parallel database system, it would be awesome to also
> parallelize some of the mapreduce phases that our data needs to go through.
> One example that just occurred to me is UNION ALL: when you union two SELECT
> statements, effectively you could run those statements in parallel. There's
> no situation (that I can think of, but I don't have a formal proof) in which
> the left statement would rely on the right statement, or vice versa. So, they
> could be run at the same time...and perhaps they should be. Or, perhaps there
> should be a way to make this happen...PARALLEL UNION ALL? PUNION ALL?
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
