September Effect

Mother's day is the busiest day for the U.S. voice telephone network.

The Friday after Thanksgiving is the busiest day for U.S. public libraries.

U.S. electrical demand forecast increased 3% this summer compared to last year.

The movie Titanic reach $600 million dollars domestic box office last

The Internet has the September effect, maybe.

Generally the previous numbers come from self-reported numbers from
companies in the various industries. The NSFNET used to report total
traffic growth. After the NSFNET was shutdown, the NAPs reported
aggregate traffic at those points. Although there are obvious problems
with any single point measurements of the Internet, they did give everyone
a gross starting point. Much of our collective notion about the hyper-growth
of the Internet came from these numbers. But if you look at the public
numbers for Internet traffic, its still growing but at a slower rate.

So what is going on?

Internet growth has slowed.

Internet growth is being constrained by some factor. If you add up
the numbers at the NAPs, there is currently no single backbone in
existence able to carry the entire Internet load. Even the new
OC-48 networks announced by AGIS and Sprint are too small.

Internet growth continues at its previous rate, but the public measurements
no longer capture it. Do overall measurements of the Internet serve any
useful purpose either for network engineering or investor information. Most
other industries report various quantitative metrics on a regular basis.

SEAN@SDG.DRA.COM (Sean Donelan) writes:

So what is going on?

Internet growth has slowed.

Internet growth is being constrained by some factor. [...]

The most recent doubling, and most of the one before it, was "the fad thing."
CB radio had a similar pair of doublings, but ultimately it got back to the
point where only people who had a use for it, used it.

The Internet is more useful than CB radio, but on the other hand when you
consider that pornography is more than half of the non-interactive traffic
(according to the log files I gathered from a worldwide product rollout in
the transparent caching space) it's not clear how many more times we'll need
to double the bandwidth of the backbone as the fad effect drops off.

So perhaps the factor constraining the Internet's growth is "good taste."

Unlike CB, the Internet provides many different services. Each
service has a "natural" amount of bandwidth that it consumes. Email
can probably get by with 1000bit/sec/user or less. DVD wants on the
order of 7-10Mbit/sec/user. Remote harddrives want 100Mbit/sec/user
or more. Tying computers together into supercomputers works better
if you have 1Gbit/sec/computer or more (depending on the app of course).

Consumer grade computers are now fast enough to handle DVD video.
CMOS camera chips are dirt cheap. So the natural course of
events would be for people to start using the Internet to download
DVD clips as well as setting up bidirectional video links.

They don't because the network connection that they can buy is
too slow to make this practical. So yes, Internet growth is