hama-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Edward J. Yoon" <edwardy...@apache.org>
Subject Re: [jira] [Commented] (HAMA-642) Make GraphRunner disk based
Date Fri, 28 Sep 2012 22:38:51 GMT
> - Does this fail always or just sometimes?

Always
> - When it finishes, is the result wrong? Just curios, how do you compare 20gb of text
files?;D

Never finishes.

> - In case it is really the combiner, does pagerank work without problems?

Never finishes if input is large.

Sent from my iPad

On Sep 29, 2012, at 5:07 AM, "Thomas Jungblut (JIRA)" <jira@apache.org> wrote:

> 
>    [ https://issues.apache.org/jira/browse/HAMA-642?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13465866#comment-13465866
] 
> 
> Thomas Jungblut commented on HAMA-642:
> --------------------------------------
> 
> A race is not good. We have to investigate a bit deeper I guess. I don't think that there
is a concurrency problem inside of jdbm, but I will have a look, maybe there is some resources
that is static, however each task has its own mutal exclusive "database". so I don't see a
problem there. 
> 
> My first guess was the use of the combiner. So here my questions:
> - Does this fail always or just sometimes?
> - When it finishes, is the result wrong? Just curios, how do you compare 20gb of text
files?;D
> - In case it is really the combiner, does pagerank work without problems?
> 
> I will build a smaller cluster in near future to test these things more efficiently.
> 
>> Make GraphRunner disk based
>> ---------------------------
>> 
>>                Key: HAMA-642
>>                URL: https://issues.apache.org/jira/browse/HAMA-642
>>            Project: Hama
>>         Issue Type: Improvement
>>         Components: graph
>>   Affects Versions: 0.5.0
>>           Reporter: Thomas Jungblut
>>           Assignee: Edward J. Yoon
>>        Attachments: HAMA-642_unix_1.patch, HAMA-642_unix_2.patch, HAMA-scale_1.patch,
HAMA-scale_2.patch, HAMA-scale_3.patch, HAMA-scale_4.patch
>> 
>> 
>> To improve scalability we can improve the graph runner to be disk based.
>> Which basically means:
>> - We have just a single Vertex instance that get's refilled.
>> - We directly write vertices to disk after partitioning
>> - In every superstep we iterate over the vertices on disk, fill the vertex instance
and call the users compute functions
>> Problems:
>> - State other than vertex value can't be stored easy
>> - How do we deal with random access after messages have arrived?
>> So I think we should make the graph runner more hybrid, like using the queues we
have implemented in the messaging. So the graphrunner can be configured to run completely
on disk, in cached mode or in in-memory mode.
> 
> --
> This message is automatically generated by JIRA.
> If you think it was sent incorrectly, please contact your JIRA administrators
> For more information on JIRA, see: http://www.atlassian.com/software/jira

Mime
View raw message