A packet or streams layer for GENSEC/SASL?

Jeremy Allison jra at samba.org
Tue Feb 7 01:29:56 GMT 2006


On Sun, Feb 05, 2006 at 06:31:18PM -0500, David Collier-Brown wrote:
> Jeremy Allison wrote:
> >  Can you explain further what you mean by this,
> >pointing to specific parts of the code please ?
> >
> >Jeremy.
> 
>   This was a comment about our previous discussions re
> filesystem performance. Unless one is using your
> large-write-buffers code, Samba both reads and writes
> smallish amounts of data from the underlying filesystem.
> 
>   In a benchmark test, large buffers yielded a startling
> improvement in read as well as write performance, even
> on single slow disks.
> 
>   This was especially visible on qfs, with multiple disks
> which is more sensitive to writes being a multiple of the
> stripe size than our ufs, which is a BSD-derived filesystem.

Ah, I thought that might be the case. Yes, we take the
client request size and only coalesce it if the write
cache is enabled. "By design" I think is the answer here.

Just one reason we can't deprecate the write cache code.

Jeremy.


More information about the samba-technical mailing list