[
https://issues.apache.org/jira/browse/HBASE-15077?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15088585#comment-15088585
]
Anoop Sam John commented on HBASE-15077:
----------------------------------------
[~tedyu]
Only int write is what we need now. As if we need later, we can add then. These
are our private interface.
ByteBufferSupportingOutputStream seems better name. ok
Performance numbers directly taking is bit difficult. Because this will be an
issue for compaction only now. That too when that compacted file block comes
from offheap L2 cache. Later on when we are with write path offheap, it will
be more impacting and always in flushes too.
[~ram_krish]
Yes we have 2 stream here. It was like a java io ByteArrayOS wrapped by a
DataOS. We need to pass a DataOS only to other areas like DBE etc.
So here I have created both versions with supporting BB apis.
Within new DataOS, we have again another stream to which we have to write BB
bytes and so we have to call the BBUtil.
[[email protected]]
Sorry I forgot to copy the details from that jira. Let me do now.
No only that line which calc the Max array size is copied. We do have same
line with same comment in another class also.I just copied from that :-)
bq.Elsewhere we say ByteBufferedCell, etc. Why then do
ByteBufferSupportedDataOutputStream and not ByteBufferedDOS?
When we say ByteBuffered it is implicit that it is backed by a BB. Here it is
different. ByteBufferSupportedDataOutputStream may not be backed by BB it
might be backed by array only. But it supports the BB write APIs. Am I making
a point?
You want to me write a test to make sure it works as we expect? Not sure how.
Any suggestion?
> Support OffheapKV write in compaction with out copying data on heap
> -------------------------------------------------------------------
>
> Key: HBASE-15077
> URL: https://issues.apache.org/jira/browse/HBASE-15077
> Project: HBase
> Issue Type: Sub-task
> Components: regionserver, Scanners
> Reporter: Anoop Sam John
> Assignee: Anoop Sam John
> Fix For: 2.0.0
>
> Attachments: HBASE-15077.patch
>
>
> HBASE-14832 is not enough to handle this. Doing the remaining needed here.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
