On Tuesday, March 25, 2003, at 11:48  AM, Ted Unangst wrote:

> I'm not positive this is still the case, but because NFS syncs data 
> after
> each write, it doesn't let LFS build a full segment.  In some 
> benchmarks
> done a while ago, this totally killed write performance.  Apart from 
> any
> stability issues.

This should not be the case with NFSv3, in which the client sends an 
explicit "commmit" request, right?

         -- Jason R. Thorpe <thorpej@wasabisystems.com>