Effect of wide links parameter.

[Christopher R. Hertel]

They obviously did a lot of work to un-tune the Linux system.  They
probably did a lot of work to speed up the NT box (Jeremy already pointed
out that they reformatted the disk to prefent trouble with fragmentation). 
The wide links parameter is just one issue (and Herb's tests were done on
Irix, not Linux). 

The other thing that Mindcruft said was that it was hard to tune Linux 
because of the "lack of information".  We can only assume that they've 
already spent years learning to work with NT.

Chris -)-----

> On Thu, 15 Apr 1999, Jeremy Allison wrote:
> 
> > Hi all,
> > 
> > 	Herb here at SGI just did a quick NetBench
> > run to determine the effects of setting the
> > "wide links" parameter to "no".
> > 
> > The result - a 30% drop in performace on NetBench
> > throughput on Samba :-).
> 
> only 30%?  what happened to the other... 45 to 50% performance difference
> between all other netbench tests and the mindcraft one?  you reckon it
> could be the file handles / mem cache thing?
> 
> luke
> 


-- 
             -- I have a shoehorn, the kind with teeth. --
                                  ---
Christopher R. Hertel -)-----                   University of Minnesota
crh at nts.umn.edu              Networking and Telecommunications Services