You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@bigtop.apache.org by "jay vyas (JIRA)" <ji...@apache.org> on 2014/05/22 17:11:02 UTC
[jira] [Commented] (BIGTOP-1270) BigPetStore: Productionize the
Hive portion
[ https://issues.apache.org/jira/browse/BIGTOP-1270?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14005995#comment-14005995 ]
jay vyas commented on BIGTOP-1270:
----------------------------------
After looking more into this : I think the hive tooling can simply be replaced with a hive script, which we maintain, but don't run as part of the official pipeline. The reason being , that hive adds a huge testing burden to bigpetstore, given the complexity of maintaining a local hive and hadoop tarball on developer machines.
Then we can wait for HIVE-1776 to come out , and once it does, bigtop can integrate it.
> BigPetStore: Productionize the Hive portion
> -------------------------------------------
>
> Key: BIGTOP-1270
> URL: https://issues.apache.org/jira/browse/BIGTOP-1270
> Project: Bigtop
> Issue Type: New Feature
> Components: Blueprints
> Affects Versions: backlog
> Reporter: jay vyas
>
> The hive portion of the BigPetStore blueprint app builds a "view" over the cleaned data that Mahout can then use to do product recommendations.
> The hive code in bigpetstore only runs locally - lets add the necessary configuration hooks and/or (if we have to) externalize the hive script itself from java so that its easy to run directly on a cluster.
> And lets actually run it on some kind of a cluster at scale. The contract for the hive portion is an output file with three numbers like this:
> {noformat}
> 100 30021 1
> 100 212341 1
> ...
> {noformat}
> Signifying that customer=100 likes both of the products "30021" and "212341".
--
This message was sent by Atlassian JIRA
(v6.2#6252)