[
https://issues.apache.org/jira/browse/HADOOP-17125?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17184686#comment-17184686
]
Taro L. Saito edited comment on HADOOP-17125 at 8/25/20, 7:07 PM:
------------------------------------------------------------------
I'm the creator of snappy-java and happy to see this change.
A couple of thoughts:
* snappy-java has pre-compiled binaries of Snappy for various platforms, which
can be found from here
[https://github.com/xerial/snappy-java/tree/master/src/main/resources/org/xerial/snappy/native]
* The PR uses only Snappy.compress, uncompress. This is a good idea as these
APIs have not been changed for years and most stable ones.
SnappyInput/OutputStreams uses snappy-java's own format, so sticking to
Snappy.compress/uncompress would be ideal for the maintainability.
* A limitation of pure-java Snappy is that it doesn't support some big-endian
architecture like IBM zOS/s390x
[https://github.com/xerial/snappy-java/issues/232.|https://github.com/xerial/snappy-java/issues/232]
Once we added a pre-compiled native library for it, we can support this
platform as well. If supporting a wide variety of big-endian platforms will be
necessary, we need to extend the aircompressor to support big-endian
architecture.
was (Author: taroleo):
I'm the creator of snappy-java and happy to see this change.
A couple of thoughts:
* snappy-java has pre-compiled binaries of Snappy for various platforms, which
can be found from here
[https://github.com/xerial/snappy-java/tree/master/src/main/resources/org/xerial/snappy/native]
* The PR uses only Snappy.compress, uncompress. This is a good idea as these
APIs have not been changed for years and most stable ones.
SnappyInput/OutputStreams uses snappy-java's own format, so sticking to
Snappy.compress/uncompress would be ideal for the maintainability.
* A limitation of pure-java Snappy is that it doesn't support some big-endian
architecture like IBM zOS/s390x
[https://github.com/xerial/snappy-java/issues/232.|https://github.com/xerial/snappy-java/issues/232]
Once we added a pre-compiled native library for it, we can support this
platform as well. If supporting more wide varieties of big-endian platform will
be necessary, we need to extend the aircompressor to support big-endian
architecture.
> Using snappy-java in SnappyCodec
> --------------------------------
>
> Key: HADOOP-17125
> URL: https://issues.apache.org/jira/browse/HADOOP-17125
> Project: Hadoop Common
> Issue Type: New Feature
> Components: common
> Affects Versions: 3.3.0
> Reporter: DB Tsai
> Priority: Major
>
> In Hadoop, we use native libs for snappy codec which has several
> disadvantages:
> * It requires native *libhadoop* and *libsnappy* to be installed in system
> *LD_LIBRARY_PATH*, and they have to be installed separately on each node of
> the clusters, container images, or local test environments which adds huge
> complexities from deployment point of view. In some environments, it requires
> compiling the natives from sources which is non-trivial. Also, this approach
> is platform dependent; the binary may not work in different platform, so it
> requires recompilation.
> * It requires extra configuration of *java.library.path* to load the
> natives, and it results higher application deployment and maintenance cost
> for users.
> Projects such as *Spark* and *Parquet* use
> [snappy-java|[https://github.com/xerial/snappy-java]] which is JNI-based
> implementation. It contains native binaries for Linux, Mac, and IBM in jar
> file, and it can automatically load the native binaries into JVM from jar
> without any setup. If a native implementation can not be found for a
> platform, it can fallback to pure-java implementation of snappy based on
> [aircompressor|[https://github.com/airlift/aircompressor/tree/master/src/main/java/io/airlift/compress/snappy]].
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
