hama-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Thomas Jungblut (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HAMA-642) Make GraphRunner disk based
Date Sat, 15 Sep 2012 18:44:07 GMT

    [ https://issues.apache.org/jira/browse/HAMA-642?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13456465#comment-13456465
] 

Thomas Jungblut commented on HAMA-642:
--------------------------------------

We could always do the hardcore database approach and save the keys in a B+ tree while keeping
the memory consuming parts on _local_ disk. 

In HAMA-596 I've seen the adjacency list use most of the RAM. So it would totally make sense
to check this approach, it would enable large undirected or even dense graphs.
Although I think it is not difficult to write a B+ Tree, I think the layout on disk and lookup
techniques could be quite sophisticated. Can we loot out some frameworks that can help us
on that?

Please don't come up with HBase or Accumulo, we should get this done without any other infrastructure
burden (at least Hama itself is enough infra burden for most people).
                
> Make GraphRunner disk based
> ---------------------------
>
>                 Key: HAMA-642
>                 URL: https://issues.apache.org/jira/browse/HAMA-642
>             Project: Hama
>          Issue Type: Improvement
>          Components: graph
>            Reporter: Thomas Jungblut
>
> 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