[
https://issues.apache.org/jira/browse/HBASE-15264?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15149763#comment-15149763
]
Duo Zhang commented on HBASE-15264:
-----------------------------------
There are several advantages
1. The fan out mechanism. This will reduce the latency.
2. The asynchronous WAL could also run in the same EventLoop, we could just
call write and flush inside the EventLoop thread, so generally we only have one
thread to do all the things. {{FSHLog}} have one disruptor consumer thread and
five sync threads default.
3. Fail-fast when connection to datanode error. The WAL implementation could
open new writer ASAP. The recovery mentioned above is only used to make closing
a little faster, does not intend to recover the writer.
4. We may benefit from the {{ByteBufAllocator}} of netty.
And the stream is designed for implementing WAL, so the idea itself has no
obvious downsides. And the current implementation, too many reflections maybe a
problem...
Thanks.
> Implement a fan out HDFS OutputStream
> -------------------------------------
>
> Key: HBASE-15264
> URL: https://issues.apache.org/jira/browse/HBASE-15264
> Project: HBase
> Issue Type: Sub-task
> Reporter: Duo Zhang
> Assignee: Duo Zhang
> Attachments: HBASE-15264-v1.patch, HBASE-15264-v2.patch,
> HBASE-15264.patch
>
>
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
