Windows XP vs Samba and asynchronous IO

[David Collier-Brown]

  If you can capture the packets with ethereal, you can
measure the latency and transfer time and confirm/deny that.
  I have scripts for doing so for http on Solaris, so if
you get a capture, email it to me and I'll have a look.

--dave

James Roper wrote:
> Hey guys,
> 
> I've been doing some experimentation with asynchronous read routines in 
> the Linux CIFS VFS client.  I've come across some interesting things 
> when comparying Windows XP to Samba that I thought you guys might be 
> able to explain.
> 
> When I first did my tests, I was copying a 173MB file from my laptop to 
> my desktop, they were linked directly with a crossover cable on a 100mb 
> network and had a 0.2 millisecond ping time.  My laptop was running 
> Linux 2.6 with Samba 3.  These are my results:
> Test *      Trial 1     Trial 2     Trial 3    Average
> Async      19.711      19.723      18.977      19.470
> Sync        21.809      21.105      21.033      21.315
> So there was about a 7-8% speed up.  Since then, the harddrive on my 
> laptop died, so I replaced the harddrive and installed Windows XP on 
> it.  It now has a 0.5 millisecond ping time in the same environment.  
> So, doing the same test, I find that using synchronous read routines 
> it's averaging about 1 minute 10 seconds, and with asynchronous routines 
> it's averaging 24 seconds, a 60-70% speed up.  Does this sound 
> right/surprise you at all?  My guess is that the Windows XP network 
> stack is very slow, it can handle high bandwidths but has a really high 
> latency.  Any other explainations?
> 
> James
> 

-- 
David Collier-Brown,      | Always do right. This will gratify
Sun Microsystems, Toronto | some people and astonish the rest
davecb at canada.sun.com     |                      -- Mark Twain