You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@thrift.apache.org by "Bryan Duxbury (JIRA)" <ji...@apache.org> on 2010/03/17 17:50:27 UTC

[jira] Updated: (THRIFT-719) Update Nonblocking and HsHa server to avoid an extra buffer copy

     [ https://issues.apache.org/jira/browse/THRIFT-719?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Bryan Duxbury updated THRIFT-719:
---------------------------------

    Attachment: thrift-719.patch

Here's a patch that I believe makes this change, but I haven't had a chance to whip up a test yet.

> Update Nonblocking and HsHa server to avoid an extra buffer copy
> ----------------------------------------------------------------
>
>                 Key: THRIFT-719
>                 URL: https://issues.apache.org/jira/browse/THRIFT-719
>             Project: Thrift
>          Issue Type: Improvement
>    Affects Versions: 0.3
>            Reporter: Bryan Duxbury
>            Assignee: Bryan Duxbury
>             Fix For: 0.3
>
>         Attachments: thrift-719.patch
>
>
> Nonblocking and HsHa servers are currently built in such a way that the memory buffer transport that actually contains the whole read message will be wrapped in a new TFramedTransport, which means the first time that you read from this transport, you'll end up copying the whole buffer again. 
> I'd like to just use a straight TMemoryInputTransport, and stop offering the option of specifying an input transport factory. Doing so is simple, but it has one side effect: you wouldn't be able to customize that transport through use of a Transport Factory. However, I think that it's just too likely someone will screw up either the functionality or the performance of the server by doing so.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.