[
https://issues.apache.org/jira/browse/HBASE-15756?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15272141#comment-15272141
]
Yu Li commented on HBASE-15756:
-------------------------------
[~aoxiang] is my workmate so allow me to give some supplement here (it seems
you forgot to add me into subscribers buddy [~aoxiang] and I just noticed it...
:-))
As found in HBASE-15619, read against empty table shows perf regression on RPC
layer, so we are trying to improve this part. And since 2.0 is still way to go,
we chose to start from trying netty to replace our current RpcServer, and got
the number [~aoxiang] posted above.
It's really good to know your thoughts [~stack], [~ram_krish] and [~apurtell],
and thanks for sharing the information that we even see a better performance
than netty when using ByteBuffers (Unsafe way of accessing), but I'm still
wondering whether there's other parts that netty does good besides ByteBufs,
and we will dig deeper into this part (and please let us know if you already
did some and had a conclusion, thanks).
Meanwhile, if offheap stuff won't be available in branch-1 and NettyRpcServer
truly does better than our current native one, maybe it's still useful for
branch-1 users? Mind share your thoughts sir [~stack] [~apurtell] [~ram_krish]
[~anoopsamjohn]? Thanks.
> Pluggable RpcServer
> -------------------
>
> Key: HBASE-15756
> URL: https://issues.apache.org/jira/browse/HBASE-15756
> Project: HBase
> Issue Type: Improvement
> Reporter: binlijin
> Assignee: binlijin
>
> Current we use a simple RpcServer, and can not configure and use other
> implementation.This issue is to make the RpcServer pluggable, so we can make
> other implementation for example netty rpc server. Patch will upload laterly
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
