You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@mahout.apache.org by "Han Hui Wen (JIRA)" <ji...@apache.org> on 2010/08/13 14:52:17 UTC
[jira] Commented: (MAHOUT-474) Should compress output of Job
pairwiseSimilarity and Job asMatrix
[ https://issues.apache.org/jira/browse/MAHOUT-474?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12898215#action_12898215 ]
Han Hui Wen commented on MAHOUT-474:
-------------------------------------
I have done test for this,
Before using patch,the output of pairwiseSimilarity is 18,7G,
After using patch ,the output of pairwiseSimilarity is about 3.952G( = 247M *16),
{code}
+ SequenceFileOutputFormat.setCompressOutput(pairwiseSimilarity, true);
+ SequenceFileOutputFormat.setOutputCompressorClass(pairwiseSimilarity, GzipCodec.class);
+ SequenceFileOutputFormat.setOutputCompressionType(pairwiseSimilarity, CompressionType.BLOCK);
{code}
Also because RowSimilarityJob run a in separated process ,the properties will lost from the main Job.
The properties must be set by the RecommenderJob
> Should compress output of Job pairwiseSimilarity and Job asMatrix
> -----------------------------------------------------------------
>
> Key: MAHOUT-474
> URL: https://issues.apache.org/jira/browse/MAHOUT-474
> Project: Mahout
> Issue Type: Improvement
> Components: Collaborative Filtering
> Affects Versions: 0.4
> Reporter: Han Hui Wen
> Assignee: Sean Owen
>
> !https://issues.apache.org/jira/secure/attachment/12451985/RowSimilarityJob-CooccurrencesMapper-SimilarityReducer.jpg!
> From above picture ,we can see that the output of pairwiseSimilarity is very large ,we should compress them.
> SequenceFileOutputFormat.setOutputCompressionType(job, style);
> SequenceFileOutputFormat.setCompressOutput(job, compress);
> SequenceFileOutputFormat.setOutputCompressorClass(job, codecClass)
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.