hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Jesse Yates (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-9534) Short-Circuit Coprocessor HTable access when on the same server
Date Mon, 16 Sep 2013 23:38:51 GMT

    [ https://issues.apache.org/jira/browse/HBASE-9534?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13768958#comment-13768958
] 

Jesse Yates commented on HBASE-9534:
------------------------------------

[~saint.ack@gmail.com] yeah, I wasn't a big fan of making those interfaces public. Alternative
was throw an UnspportedOperationException if you try to hit the master with that connection.
Keeps the interfaces how they are, but given that we will have this connection, we could just
as easily use it to talk locally to the master (though, as I mention in the description, its
not really worth it).

I'm ok rolling those back and going with the exception or marking them InterfaceAudience.Private
- former seems cleaner for the moment.
                
> Short-Circuit Coprocessor HTable access when on the same server
> ---------------------------------------------------------------
>
>                 Key: HBASE-9534
>                 URL: https://issues.apache.org/jira/browse/HBASE-9534
>             Project: HBase
>          Issue Type: Bug
>            Reporter: Jesse Yates
>            Assignee: Jesse Yates
>              Labels: coprocessors, performance, regionserver
>             Fix For: 0.98.0
>
>         Attachments: hbase-9534-trunk-v0.patch
>
>
> Coprocessors currently create a full HTable when they want to write. However, we know
that coprocessors must run from within an HBase server (either master or RS). For the master,
its rare that we are going to be doing performance sensitive operations, but RS calls could
be very time-intensive. 
> Therefore, we should be able to tell when a call from a CP attempts to talk to the RS
on which it lives and just short-circuit to calling that RS, rather than going the long way
around (which does the full marshalling/unmarshalling of data, as well as going over the loopback
interface).

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

Mime
View raw message