giraph-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Jan van der Lugt (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (GIRAPH-259) TestBspBasic.testBspPageRank is broken
Date Wed, 01 Aug 2012 16:18:03 GMT

    [ https://issues.apache.org/jira/browse/GIRAPH-259?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13426726#comment-13426726
] 

Jan van der Lugt commented on GIRAPH-259:
-----------------------------------------

I'm sorry, but I don't agree to the points made by Alessandro. MasterCompute _should_ be executed
before all other vertices are. I don't know if this breaks the BSP model, since it simply
creates another barrier. This will, however, not result in a scalability issue, since there
will always be at most one MasterCompute. By having the MasterCompute execute before other
vertices, it enables a more intuitive implementation of algorithms like betweenness centrality,
which have different phases. The MasterCompute sets the current phase in a variable and all
the vertices use this variable to determine what they will do. The only way for MasterCompute
to communicate with other vertices should be aggregators: sending messages to every node would
be very costly. Stanford GPS also uses this model (interleaving vertex and master computation),
but doesn't rely on aggregators, but instead on a global objects map. They are functionally
the same. If MasterCompute would run together with the vertices, there would be supersteps
where the vertices would be idle, because they are waiting for a value from MasterCompute,
which would be very wasteful.

Since MasterCompute depends so heavily on aggregators, given they are its sole means of communication,
I don't think collecting aggregators and running MasterCompute are completely orthogonal components.

Hope I made this clear enough, otherwise let me know.
                
> TestBspBasic.testBspPageRank is broken
> --------------------------------------
>
>                 Key: GIRAPH-259
>                 URL: https://issues.apache.org/jira/browse/GIRAPH-259
>             Project: Giraph
>          Issue Type: Bug
>            Reporter: Maja Kabiljo
>            Assignee: Maja Kabiljo
>         Attachments: GIRAPH-259-1.patch, GIRAPH-259-2.patch, GIRAPH-259-3.patch
>
>
> Test crashes on line 152 in class SimplePageRankVertex in distributed mode.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira

       

Mime
View raw message