[plug] modem connect statistics

[Bret Busby]

Matt Kemner wrote:
> 
> Out of the 62 people currently connected to WASP, one logged in at
> 52000 baud, 25 logged in at 50666.
> 
> 56k modems retrain up and down constantly though. Most of those have
> retrained down to 46666 or so.
> 
> Of the 62 people logged in, 8 are currently at 50666, 20 are between that
> and 46666, 19 between that and 41333, 0 above 33600 and below
> 41333, leaving 15 at 33600 and below (and all of those are connected via
> V34, the rest are using either K56 or V90)
> 
> So everyone currently connected to wasp with a 56k capable modem is
> connected at least at 41333, and 60% of those are at 46666 and above
> 
> Interesting the kind of statistics one can get from digital modems
> 
>  - Matt

Thanks for the statistics, Matt. They are very useful. But, one point;
could you advise, apart from the connection speeds (capability, or,
potential), what the transmission speeds (actual) are like? There are
three figures here, I suppose; capability figures, and, one actual
figure. The two capability figures, are the modem speeds and the connect
speeds (the capability of the modem, and the capability of a stable
connection, using the modem). The actual figure, is the download speed,
being the speed at which the data downloads, given the two capability
figures (limited by the connect speed), and, limited by all kinds of
other factors, that are independent of the connect speed (such as
congestion at the ISP, and, congestion and traffic load, at every point
from there to the websites being accessed, and back again).

It may sound like quibbling, but, going back to my previous point about
the lamborghini and the narrows bridge, to obtain connection speed of a
certain rate is one thing, but, is there any real advantage in the
transmission rates, between the two modem speeds (of 33600 and 56k)?

Do your stats give the variance of the transmission (especially, the
download) rates? If for example, statistics could be given, like a mean
speed of 41333, with a standard deviation of 2000, would mean that 68%
of the conected people were achieving download speeds of between 39333
and 43333 bits (or bytes, as the case may be) per second, which is the
important figure. If not a mean speed, with standard deviations, then a
median speed, with percentiles, so, for example, 60% achieving download
speeds of between 39000 and 43000, with 50% achieving speeds greater
than 41000, etc.

That is even more useful; the actual transmission speed ranges. If you
can give that,it would be great. Of course, that is probably something
that would need to done over a time period, for example a day, or a
week, or a month, rather than an instantaneous snapshot, due to
fluctuations in traffic and congestion.

Of course, to balance that, instead of just having the variance of the
download speeds, the variance of the connect speeds for the same period,
would also need to be included. There wouldn't be much point, for
example, in indicating that download speeds beyond the capability of a
33600 modem, were only obtained by 25% of the people connected, over a
period, if, over that period, only 25-30% of the connections, were at a
speed greater than 33600.

-- 

Bret Busby

......................................