[
https://issues.apache.org/jira/browse/ORC-201?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16027078#comment-16027078
]
Gopal V edited comment on ORC-201 at 5/27/17 12:57 AM:
-------------------------------------------------------
Brotli is clearly aimed at performing better at longform ASCII Text data.
See Appendix B transforms (and decode Appendix A) -
https://tools.ietf.org/html/rfc7932#appendix-B
The hadoop benchmark does not naturally translate into a speedup for ORC as
Hadoop Gzip defaults to Zlib-6 (which is deflate_slow), while Orc sticks to
only deflate_fast (Zlib 1,2,3).
was (Author: gopalv):
Brotli is clearly aimed at performing better at longform ASCII Text data.
See Appendix B transforms (and decode Appendix A) -
https://tools.ietf.org/html/rfc7932#appendix-B
The hadoop benchmark does not naturally translate into a speedup for ORC as
Hadoop Gzip defaults to Zlib-6 (which is deflate_slow), while Orc sticks only
deflate_fast (Zlib 1,2,3).
> Brotli compression codec support
> --------------------------------
>
> Key: ORC-201
> URL: https://issues.apache.org/jira/browse/ORC-201
> Project: ORC
> Issue Type: New Feature
> Components: compression
> Reporter: Prasanth Jayachandran
>
> HADOOP-13126 is bringing Brotli compression codec to hadoop. ORC should add
> support for Brotli as it seems to have better performance that SNAPPY.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
