You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@reef.apache.org by "Markus Weimer (JIRA)" <ji...@apache.org> on 2015/09/28 00:08:04 UTC
[jira] [Commented] (REEF-788) IMRU Map and Update host tasks take a
lot of internal memory
[ https://issues.apache.org/jira/browse/REEF-788?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14909910#comment-14909910 ]
Markus Weimer commented on REEF-788:
------------------------------------
Keeping that data around could be a good idea, couldn't it? As soon as we add fault handling to IMRU, we will have to do something very similar.
> IMRU Map and Update host tasks take a lot of internal memory
> ------------------------------------------------------------
>
> Key: REEF-788
> URL: https://issues.apache.org/jira/browse/REEF-788
> Project: REEF
> Issue Type: Sub-task
> Components: REEF.NET
> Affects Versions: 0.13
> Environment: C#
> Reporter: Dhruv Mahajan
> Assignee: Dhruv Mahajan
> Fix For: 0.13
>
>
> Currently in map and update host tasks in IMRU input and output are maintained in memory between the iterations. Also for large sized messages a lot of internal memory is being utilized for pipelining.
> The aim of the jira is to remove the maintenance of input and output between the iterations. Also GC.Collect is used for now to remove memory overheads.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)