Mailing-List: contact dev-help@hama.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@hama.apache.org
Date: Tue, 8 Sep 2015 05:39:45 +0000 (UTC)
From: "ChiaHung Lin (JIRA)" <jira@apache.org>
To: dev@hama.apache.org
Message-ID: <JIRA.12862294.1441677588000.275884.1441690785739@Atlassian.JIRA>
In-Reply-To: <JIRA.12862294.1441677588000@Atlassian.JIRA>
References: <JIRA.12862294.1441677588000@Atlassian.JIRA>
 <JIRA.12862294.1441677588693@arcas>
Subject: [jira] [Commented] (HAMA-974) Support fault tolerance for Graph job
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit


    [ https://issues.apache.org/jira/browse/HAMA-974?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14734263#comment-14734263 ] 

ChiaHung Lin commented on HAMA-974:
-----------------------------------

As we discussed previously, the main difference is to provide superstep as the unit for computation so the the previous messages, etc. can be put back in iteration. Other changes are centered on this issue. 

I've merge trunk to the test-sched branch which is the latest one so you can compare the diff. 


> Support fault tolerance for Graph job
> -------------------------------------
>
>                 Key: HAMA-974
>                 URL: https://issues.apache.org/jira/browse/HAMA-974
>             Project: Hama
>          Issue Type: Improvement
>          Components: graph
>    Affects Versions: 0.7.0
>            Reporter: Edward J. Yoon
>             Fix For: 0.8.0
>
>
> Currently we only checkpoints messages. To support FT for graph job, aggregators, assigned vertices and its statuses must be checkpointed together.


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