[ILUG] Re: [IIU] volunteers wanted for IIU survey

[Alex French]

At 6:43 PM +0100 18/4/2000, Justin Mason wrote:

>
>It does neglect MRU/MTU settings, modem speed, latency, and number of dial
>attempts before a connection is established; mostly these are neglected
>because I'm not sure how to get that data portably or just haven't done it
>yet...


Fair enough! :-)

>(also I'm not entirely clear how the MTU thing will work, I thought that
>was a missing feature in MS TCP/IP from Windows 95, and modern TCP/IP
>impls support MTU discovery?  That's what I was assuming anyway.)

Some do, but MTU discovery is fraught with Issues(tm), not least of 
which are overagressive firewalls.....

>
>But the main idea is that it samples the PPP statistics every 5 seconds,
>and therefore can be used to get an idea of the peak throughputs for
>bursts of traffic on that PPP session.  From that, a rough idea of the
>highest data rate can be correlated to the ISP.

But that could be an ftp session from the ISP's own server, or one 
particularly well-connect site? If that's the case, this really just 
tells us about the ISP's modems & internal network.

How about this as a strategy: pick 10 sites (at major exchanges, say) 
and have the software do periodic pings to them. Although this is far 
from perfect (it neglects all the things I mentioned above) it might 
be interesting.

To do an even better test, pick one of the sites be inside the ISP's 
network, and subtract this value from the others. This will correct 
for all the factors I mentioned.

NB: this will of course only measure latency, not throughput.
-- 

Alex.