[From: "Jeffrey J. Mountin" <sysop@mixcom.com>]
> Measureable yes, but 3 100ths of a second? Compared to delays on congested
> part of the net, this is nothing. Ping times are one thing, real time
> human perception is quite another, IMHO.
>
> Number freaks.
> FWIW, when I change something here that looks good to me by the numbers I
> ask for the "how does it work for you" opinion.
[From: "Michael R. Wayne" <wayne@msen.com>]
> It doesn't really matter whether >I< think ping is a reasonable
> benchmark, my users will quit if the ping times are long. We have a
> PM3 announced in "test" mode right now, almost all the mail from our
> testers reads like:
> Hey, I got a 33.6 connect on the new number but the ping times suck.
> What's wrong with these lines?
> I'll go back to the analog lines, thanks.
> typically including a few hundred lines of ping results reflecting the
> problems they see.
I am assuming that most of you smaller ISPs actually run game
servers or have users who run on them.
As usual with games, they push whatever technology you use to the
limit. In this particular case, the traffic is highly bursty,
and a LOT of it. Ping times are one metric that the users can measure
and report to complain about, but due to the bursty nature of the
traffic, they seem to be getting more sluggish responses from their
games.
Although the actual throughput on any given sustained
transaction seems to be improved, the speed of bursty transactions
seems to be lower due to some form of increased latency.
If this info isn't sufficiently helpful, I can waste some time writing
code to hammer a connection at various packet sizes. But it'll have
to be in my "free time".
Livingston is, of course, welcome to a dialup account on the two PM's
in question to play games and find out for themselves. Contact me
for details.
-- Jeffrey Haas -+- jmh@msen.com -+- http://www.msen.com/~jmh /\/\sen, Inc. "Michigan's Best Run Internet Service Provider."