Wrong commit...

this is the culprit:

commit 8fe9a12b1ad8dc9cd35324f4ed53bea9cb37ce22
Author: Robert Gemmell <rob...@apache.org>
Date:   Fri Jun 27 10:07:47 2014 +0000

    PROTON-597: update TransportOutputAdaptor and FrameParser to release 
buffers after use, reducing memory consumption when using large frame sizes
    
    Change from Marcel Meulemans
    
    git-svn-id: https://svn.apache.org/repos/asf/qpid/proton/trunk@1606010 
13f79535-47bb-0310-9956-ffa450edef68

On Jul 2, 2014, at 11:40 AM, Clebert Suconic <csuco...@redhat.com> wrote:

> commit ccb7b03b8089c2fea329d44cea5dc06da6345ea6
> Author: Robert Gemmell <rob...@apache.org>
> Date:   Fri Jun 27 10:08:24 2014 +0000
> 
>     PROTON-597: only release transport buffers if they are above a threshold 
> capacity
>     
>     git-svn-id: https://svn.apache.org/repos/asf/qpid/proton/trunk@1606012 
> 13f79535-47bb-0310-9956-ffa450edef68
> 
> 
> 
> It makes Proton pretty much useless...
> 
> this code makes proton create a new buffer (a big buffer actually) every new 
> message written.. even minor messages.
> The previous version before this would always use the same buffer, recycling 
> it... (making copies between my buffer and Proton), and then reseting the 
> buffer ready for the next usage.
> 
> 
> with this new change in place, a benchmark that I was already struggling to 
> improve would make 2500 very small messages / second. With the new change you 
> get it to 333 small messages / second...
> 
> 
> 
> 
> If you want to diminish the weight of each connection, then you need a proper 
> refactoring to Proton where you write to output buffers rather than internal 
> buffers. Proton needs this buffer as it is now unless you do a proper 
> refactoring.

Reply via email to