[ http://issues.apache.org/jira/browse/HADOOP-538?page=all ]
Arun C Murthy updated HADOOP-538:
---------------------------------
Attachment: HADOOP-538.patch
HADOOP-538_benchmarks.tgz
Here's a patch incorporating 2 major features:
a) A framework for decoupling the '(de)compressor' from the 'stream' as
discussed on this issue i.e. the {Com|Decom}pressor interfaces and a concrete
implementation of a Data{Com|Decom}pression{Out|In}putStream for plugging the
(de)compressors.
b) A straight-forward wrapper over the popular zlib algorithm
(http://www.zlib.net) using a 'direct' java.nio.ByteBuffer - this gives us
60%-70% improvement in performance over existing java.util.zip.{De|In}flater
(benchmarks attached).
I have also added some amount of framework (tweaks in build.xml to build the
native libhadoop.so, Makefiles, src/native directory etc.) so that any future
'native' code written for hadoop will fit into the native-code framework
(src/native dir and libhadoop.so) fairly easily... I have built the
infrastructure so that only 1 share-object (libhadoop.so) is created by the
top-level Makefile (src/native/Makefile) and it should have *all* of the
necessary 'native' code for hadoop.
I'd appreciate feedback/improvements to any of the above stuff...
Doug, could you please treat this patch as 99.9% ready and hold-off committing
it (assuming you are satisfied with it) so as to get as much feedback over this
as possible? Thanks!
> Implement a nio's 'direct buffer' based wrapper over zlib to improve
> performance of java.util.zip.{De|In}flater as a 'custom codec'
> -----------------------------------------------------------------------------------------------------------------------------------
>
> Key: HADOOP-538
> URL: http://issues.apache.org/jira/browse/HADOOP-538
> Project: Hadoop
> Issue Type: Improvement
> Affects Versions: 0.6.1
> Reporter: Arun C Murthy
> Assigned To: Arun C Murthy
> Fix For: 0.7.0
>
> Attachments: HADOOP-538.patch, HADOOP-538_benchmarks.tgz
>
>
> There has been more than one instance where java.util.zip's {De|In}flater
> classes perform unreliably, a simple wrapper over zlib-1.2.3 (latest stable)
> using java.nio.ByteBuffer (i.e. direct buffers) should go a long way in
> alleviating these woes.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://issues.apache.org/jira/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
