[
https://issues.apache.org/jira/browse/HADOOP-12010?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15548746#comment-15548746
]
Kai Zheng commented on HADOOP-12010:
------------------------------------
Hi [~andrew.wang] and [~zhz],
I thought this issue is obsolete now because it's opened at the time when the
new RS Java coder wasn't developed. It's actually opened against the HDFS-RAID
coder. To address this consideration and have a Java coder that's compatible
with the native coder, a new Java coder had been developed and already
committed elsewhere.
If you'd agree, I'd close this a dup. Please comment, thanks.
> Inter-operable between Java RS erasure coder and native RS erasure coder
> ------------------------------------------------------------------------
>
> Key: HADOOP-12010
> URL: https://issues.apache.org/jira/browse/HADOOP-12010
> Project: Hadoop Common
> Issue Type: Sub-task
> Affects Versions: 3.0.0-alpha1
> Reporter: Kai Zheng
> Assignee: Kai Zheng
> Labels: hdfs-ec-3.0-must-do
>
> It's natural and desired to support inter-operable between the two
> implemented RS raw erasure coders, in other words, data encoded in one coder
> (like {{RSRawEncoder}}) can be decoded using the other coder (like
> {{NativeRSRawDecoder}}). Without this, raw erasure coder will not be
> transparent to HDFS data. As such support isn't trivial and involves
> challenge work because the two implementations use different encode/decode
> matrix generation algorithms, better to have this separately.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
