[
https://issues.apache.org/jira/browse/TINKERPOP-964?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15087625#comment-15087625
]
Marko A. Rodriguez commented on TINKERPOP-964:
----------------------------------------------
Excellent! Can you provide some details on your experiments if you have them?
For instance:
* How many machines in the cluster?
* How big was the dataset (vertices and edges)?
* What OLAP jobs did you run?
* Any configuration issues/stumbling-blocks when trying to get this working?
* Any timing/performance statistics if you have them?
** The primary point of this ticket was to ensure compatibility and not
benchmarking, but if you have some stats, that would be really cool.
Finally, if you are confident that GiraphGraphComputer and SparkGraphComputer
both work correctly on a real Hadoop2 cluster (non-pseudocluster), please close
this ticket.
Thank you,
Marko.
> Test XXXGraphComputer on a Hadoop2 cluster (non-pseudocluster).
> ---------------------------------------------------------------
>
> Key: TINKERPOP-964
> URL: https://issues.apache.org/jira/browse/TINKERPOP-964
> Project: TinkerPop
> Issue Type: Improvement
> Components: hadoop, test-suite
> Affects Versions: 3.1.0-incubating
> Reporter: Marko A. Rodriguez
> Assignee: Jason Plurad
> Fix For: 3.1.1-incubating
>
>
> Seems that [~drobin1437] was having troubles with {{GiraphGraphComputer}} on
> his Hadoop2 cluster, where (crazy enough) it works fine on his Hadoop2
> pseudo-cluster. As such, I think we should test both {{SparkGraphComputer}}
> and {{GiraphGraphComputer}} on real clusters. Moreover, it would be great to
> test with the 1-billion edge Friendster graph. We had done this before in the
> {{MX}}-days, but we should make it part of our testing process (even if its
> manual). Perhaps using AWS to spin up a cluster "easily."
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
