[Microtimes article on why the Internet is so slow.]
Pretty basic but still kind of interesting...
microtimes.com
Excerpt: "Everybody senses a booming industry and wants to buy into it. Not everybody is going to win
in the end, but those who have deep enough pockets to keep pouring in cash for another 3-5
years, until at least the US Internet finally saturates the market, will then begin to reap their
financial rewards. And in the meantime a lot of people are playing the stock market."
Internet * By John S. Quarterman
Internet Economics 101
Why Is It Slow?
Every month or two a reporter asks me, "Why is the Internet so slow?"
Usually the next question is, "Why doesn't the Internet slow to a stop?"
To answer these questions, we have to discuss the economics of the Internet. I've alluded to
this before, most recently in "Is the Internet a Commons?" But I haven't gone into it in detail in a
column before.
Disclaimer: I am not an economist; I merely play one in this column. I'll be happy to entertain
any refinements to what I say here.
Internet Money Flow
Where does Internet capacity come from, anyway?
Users pay Internet Service Providers (ISPs) so they can dial-up an ISP. Local ISPs pay wide
area ISPs for long-haul service over leased lines. ISPs pay Network Access Points (NAPs) to
exchange traffic.
Wide area ISPs don't generally pay each other to exchange traffic, but that's another subject
that I don't want to get into right now. Call that one Internet Economics 102.
There are numerous variants on this simple scheme. An end-user may use some access to
technology other than a modem, such as ISDN, ASDL, cable modem, or satellite. The
end-user may be a person, a university, a company, or some other entity, large or small, rich or
poor. ISPs may use all sorts of different technologies. But they all use the Internet Protocol (IP)
or we wouldn't be talking about the Internet. And the money flows generally from the end-user
upwards through smaller to larger ISPs. Each Internet Service Provider uses some of that
money to pay for capacity.
There are also other sources of money. People and companies pay Web farms (which may or
may not also be ISPs) to serve Web pages to the Internet. These Web farms, or Web hosting
companies, or Web service providers, pay ISPs for Internet access, and that money feeds into
the same channels as end-user ISP access money.
People pay for information and other products and services over the Internet. The companies
that accept payment or provide delivery of such products and services pay ISPs for their own
access. That's more money in the cash river.
And there is a big complication that leads to another source of money. The Internet
approximately doubles in size every year. This is because of user demand, which is as it should
be. But growth at such a rate makes it almost impossible for an ISP to pay for necessary new
capacity out of cash flow. We're talking as many new servers, routers, and as much more
bandwidth every year as the ISP had total the previous year.
The extra money is coming from somewhere. Where? Investment. From the ISP owners, from
banks, from venture capitalists, and, especially for the big ISPs, from the stock market.
Everybody senses a booming industry and wants to buy into it. Not everybody is going to win
in the end, but those who have deep enough pockets to keep pouring in cash for another 3-5
years, until at least the US Internet finally saturates the market, will then begin to reap their
financial rewards. And in the meantime a lot of people are playing the stock market.
"Yes, but why is the Internet so slow?"
So how does tracing Internet money flows answer the question? Well, deployment of new
capacity happens at about the same rate as money flows into the Internet from users and their
traffic. Internet responsiveness being roughly capacity divided by traffic, people continue to see
the Internet as slow.
It's Actually Getting Faster
Actually, the Internet is getting faster, and has been doing so consistently over the long term for
years now. For graphical evidence of this, see our MIDS Internet Weather Report
www.mids.org/weather/, especially the longterm graphs www.mids.org/weather/pingstats/ .
Longtime Internet users don't really need these graphs to prove this point. Those of us who can
remember the 56Kbps first version of NSFNET, or the T-1 (1.544Mbps) second version, or
even the T-3 (45Mbps) third version, don't need any reminders that things are way faster now.
I know, the NSFNET was partly funded by tax money. But the commercial ISPs that started in
1991 (AlterNet and PSINet were the first two) and exploded in 1994 follow the commercial
economic model I have outlined. The commercial ISPs started mostly as T-1 networks and
quickly expanded through T-3 to even faster speeds.
A Historical Answer
The easiest way to answer the question about "Why doesn't the Internet slow to a stop?" is to
look at history.
Around 1993, the World Wide Web (WWW or the Web) started to be deployed. Mosaic
was released in April of that year, causing the first big surge of Web use.
Lots of people were worried. Transferring pictures over the Internet was unusual, and here was
an Internet service that made pictures easy. It even promoted using pictures with every
information exchange! Surely all this new traffic would overload the Internet. After all, a single
GIF image could be bigger than one person's typical day's worth of electronic mail traffic.
The perceived problem got even worse in 1995, when Netscape became popular. It not only
transferred images, it deliberately transferred four of them at a time! Surely this was user
convenience at the cost of damaging the Internet infrastructure everyone depended on?
Well, that's not the way it worked out. Instead, the Web boom attracted hordes of new users
to the Internet. These users were, for the most part, willing to pay to see these fast pictures. A
whole new industry of small ISPs sprang up to provide the access the users wanted to pay for.
The small ISPs paid bigger ISPs for access. The bigger ISPs put in more bandwidth and links
to handle the demand. The entire Internet grew, by every measure, even faster than its usual
very fast rate.
Because of this boom, the year 1994 is often mentioned reverently by those who first
discovered the Internet then, and especially by those who made their fortunes in the boom that
started then. Maybe we should invent an Internet dating system, with pre-1994 and post-1994
eras that are radically different. (I find it amusing when someone says to me, "You were on the
Internet way back before 1994?")
So more traffic paradoxically produced more bandwidth. And that wasn't an accident. This is
the way Internet economics works.
An Intuitive Answer
Recently I got a more specific question: "Why does everyone see an Internet slowdown at
about the hours of 3-5pm Eastern Time?"
I know it's considered tacky these days to bring up the tired old highway analogy, but
sometimes it still is useful.
Why do traffic jams on highways happen around 9am and 5pm? Because those are peak usage
times.
Why do traffic jams happen at all? Because the taxpayers or toll road users are not willing to
pay for enough capacity to completely prevent them. Even if they were willing to pay,
deployment of the new capacity would take a while.
Why don't traffic jams happen all the time? Because the taxpayers wouldn't put up with that.
The Internet differs from highways in that its feedback loops are much faster. Also, the kinds of
traffic the Internet carries, and the types of links it includes, vary much more quickly than for
highways. But the basic economic feedback mechanism is the same.
Just as on the Internet, the highways capacity tends to be enough for average load, but not
enough for peak load.
If the Internet had enough capacity to avoid slowdowns in the afternoon, lots of people would
move to cheaper ISPs. That would lead to less capacity (or to capacity not keeping up as
closely with traffic growth, which amounts to the same thing). Afternoon traffic slowdowns
would result.
If the Internet had so little capacity that it always seemed slow, either people would pay more
and we'd get more capacity, or people wouldn't sign up as quickly, and capacity deployment
would catch up.
A Matter Of Perception
So why do people still think the Internet is slow? Partly due to expectations.
For one thing, people don't notice gradual improvement, but they do notice sudden changes for
the worse. There are seasonal changes in Internet responsiveness. In January and September
every year, the Internet does in fact get slower over the short term. Why? All those students
going back to college.
For another thing, there is the Internet new user phenomenon. Remember, the Internet is
doubling in size each year. That means every year half of the Internet's users weren't users last
year. Where did they get their expectations, then? From television demonstrations of Internet
use, or from canned single-machine demonstrations, both of which naturally use optimal
conditions. Or from personal computer response times, which of course are wildly unrealistic as
metrics for wide area network response times.
Or are they?
The Internet has been created out of nothing because people wanted it. There are physical
limits to what can be done over computer networks, such as the speed of light. But most of the
delays that people perceive in the Internet now are not up against those limits yet. People's
perceptions will cause the Internet to improve to meet them.
Copyright c 1998 MIDS mids@mids.org, www.mids.org (512)451-7602, fax:
(512)452-0127, 1106 Clayton Lane, Ste. 500W, Austin, TX 78723, U.S.A.
| Home | Overview | About MicroTimes | Current Issue | Past Issues | Family and Friends |
| Great Mall of MicroTimes | Quotation Netline | Netsurf | Resources | Publishing Schedule | |
