[
https://issues.apache.org/jira/browse/HAMA-579?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Thomas Jungblut updated HAMA-579:
---------------------------------
Attachment: HAMA-579_final.patch
Final patch, fixed few things.
Very interesting note on this, I observed a bug Giraph also has. [1]
I refactored the aggregator usage bit and added a few check statements into
pagerank. This works now with convergence detection and global summation.
However this information is available from superstep 2, since the first
superstep is just compute with the vertex value, then there is messaging phase,
followed by a compute phase and then there is sync done and the master should
aggregate the values.
Little drawback but should be not too bad.
Please review, if you don't want to, I'm going to commit this tomorrow as usual.
BTW just like discussed in [1] should we move our aggregators to a subpackage?
[1]
https://issues.apache.org/jira/browse/GIRAPH-192?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13280909#comment-13280909
> Add multiple aggregators
> ------------------------
>
> Key: HAMA-579
> URL: https://issues.apache.org/jira/browse/HAMA-579
> Project: Hama
> Issue Type: Improvement
> Components: graph
> Affects Versions: 0.5.0
> Reporter: Thomas Jungblut
> Assignee: Thomas Jungblut
> Fix For: 0.5.0
>
> Attachments: HAMA-579_1.patch, HAMA-579_2.patch, HAMA-579_final.patch
>
>
> Very easy to implement, on client side we use varargs feature to set multiple
> classes. On job side we extend the single aggregator to a list. During the
> computation it will just loop over the list and execute the things it did.
> We have to assign the messaging of the values an index of the aggregator list.
--
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
