You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@ignite.apache.org by "Vladimir Ozerov (JIRA)" <ji...@apache.org> on 2016/12/05 13:49:58 UTC

[jira] [Commented] (IGNITE-4283) Hadoop: implement "striped" shuffle mode.

    [ https://issues.apache.org/jira/browse/IGNITE-4283?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15722309#comment-15722309 ] 

Vladimir Ozerov commented on IGNITE-4283:
-----------------------------------------

Design:
1) Every mapper works with it's own set of "remote" output maps;
2) These maps are essentially not "maps", but IO messages, which we fill up to certain threshold;
3) Once filled, message is sent to remote node.
4) Async shuffle thread is no longer need in this architecture.

As a result we decrease concurrency, removes slowdown from a single shuffle thread which is not able to send messages fast enough, and removes unnecessary intermediate sorting.

NB! Be careful with "combiner" case and with "external" execution.

> Hadoop: implement "striped" shuffle mode.
> -----------------------------------------
>
>                 Key: IGNITE-4283
>                 URL: https://issues.apache.org/jira/browse/IGNITE-4283
>             Project: Ignite
>          Issue Type: Sub-task
>          Components: hadoop
>    Affects Versions: 1.7, 1.8
>            Reporter: Vladimir Ozerov
>            Assignee: Vladimir Ozerov
>             Fix For: 2.0
>
>
> No need for this.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)