[
https://issues.apache.org/jira/browse/HAMA-807?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Anastasis Andronidis updated HAMA-807:
--------------------------------------
Status: Patch Available (was: In Progress)
I attached a patch with the feature. I tried it on an application that I have
and seems to work.
There is a limitation though:
AggregationRunner::aggregateVertex is not skipped due to the fact that we need
to add an extra superstep to determine all aggregators that need skipping.
We can make it work with only local skip indexes to gain more speed. But we
make the assumption that all vertices run the same code. (see:
http://mail-archives.apache.org/mod_mbox/hama-user/201309.mbox/%[email protected]%3e)
I suggest to add this functionality (if it works properly) and delegate
refactoring and speedup in HAMA-788.
> Make aggregators skip supersteps
> --------------------------------
>
> Key: HAMA-807
> URL: https://issues.apache.org/jira/browse/HAMA-807
> Project: Hama
> Issue Type: New Feature
> Components: graph
> Affects Versions: 0.6.2
> Reporter: Anastasis Andronidis
> Assignee: Anastasis Andronidis
> Priority: Minor
> Labels: aggregator, features, graph
> Fix For: 0.6.3
>
> Attachments: HAMA-807.patch
>
>
> Users should be able to control their aggregators better by making
> aggregators to skip calculations in specific supersteps.
> This functionality can be accessed by the method:
> skipAggrigator(int aggrIndex);
--
This message was sent by Atlassian JIRA
(v6.1#6144)
