At 08:56 PM 7/26/2004, Mitch \(WebCob\) wrote:

> > Hmmm, I tried a 1456 byte ping and see a 130ms latency without the
> > shaper on. Wow, I didn't think about doing this. That would certainly
> > make sense! Glad to see it wasn't m0n0wall after all! I guess my best
> > option would be to find a way to increase my upload.
>
>I'm not an expert at this part, but I seem to recall various tools that
>would optimize a link by changing the mtu based on what works best for the
>link. Sometimes iirc, what happens is that although YOU have an mtu of say
>1400 with your ISP, SOMEWHERE in the chain between you and them, there is a
>link with an mtu of say 614 ;-)
>
>This results in all your packets being fragmented in the middle, and
>reassembled on the other side, which can increase delays more I think...
>this is where some of that wacky windows mtu optimizing software got started
>I think.
>
>Be happy to hear from a few gurus if any are listening though.
>
>m/

You're on the right track.  All reasonably modern equipment is *supposed* 
to be capable of reporting it's MTU settings, etc., and the clients can 
find this information by doing the appropriate discovery requests.  That 
seems to work pretty well for most folks, but there are apparently some 
pieces of equipment still in the wild that either don't report, don't 
report correctly, or otherwise fall down on the job in this respect.  The 
other thing that can happen is that folks often run some of the available 
tweaking routines to optimize their settings and disable the automatic 
discovery portion of things.  It's been too long since I've played with any 
of the optimizers to remember which ones were doing that.  At the time is 
was mostly manifesting itself in the IPSec setup I was working on.

The simplest means to determine if everything is working correctly is the 
ping command with the packet size and fragmentation flags set.  That should 
tell you what's going on, at least between any 2 endpoints.  Different 
endpoints could lead to different results, but normally I've found that the 
differences are predominantly pretty close to your end of the setup so they 
don't change much after you figure out what's going on.

HTH

Melvin


----------------------------------------
Good, Fast, Cheap... Pick any two
----------------------------------------