The Forrest Curve
Jon Forrest
forrest@ce.berkeley.edu

There is a phenomenon sweeping the computer industry that is having
a profound but largely unrecognized effect. I claim that companies
ignoring this phenomenon will suffer a slow and painful death. What's
more, there's absolutely nothing that can be done to escape it.
In this article I first describe this phenomenon and then spend some
time trying to figure out what it means.

Simply and briefly stated, my hypothesis is that fewer and fewer
computer users think their computer is too slow. I've invented what I
call the Forrest Curve to illustrate this.

Here's the Forrest Curve:

| \
| \
"slow" | \ /\
factor | v \
| \ /\
| v \
| \ /\
| v \
| \
______________________

-- time --

This is a curve with a general downward slope, having occasional
upward blips. The curve approaches but never hits 0. (Neither axis
is drawn to any scale nor can be used to derive any specific numeric
values).

The "slow" factor is the number of people who think their computer
is too slow. I admit that this isn't a very objective measurement but
you can get a feel for it by the amount of grumbling about computer
speed that takes place in your office.

I'm also not being very specific about exactly what constitutes a
"computer". I claim it really doesn't matter. Taken as a
whole, the whole pile of stuff sitting on someone's desk (or lap), is
what I consider to be a computer. A more detailed examination wouldn't
change the Forrest Curve.

Note that I'm including the entire population of computer users in
this graph, many of whom know very little, if anything, about what
their computer is really doing or how it works. But, even if I were
to confine this graph to "software professionals" the graph would
merely have a higher origin point. The general shape wouldn't
change.

I also recognize that there is a class of users that can and will
always be able to consume any amount of computer resources.
These guys are why the Forrest Curve never goes to zero. In spite of
their needs they can't reshape the Forrest Curve because they don't
have enough money to spend anymore.

Every so often something comes along that causes a temporary
perturbation in the Forrest Curve. Some examples might be
relational databases, X-windows, Windows NT, multimedia,
handwriting and speech recognition, and so on. This is natural.
There will always be such cases and they admittedly can cause high
blips in the Forrest Curve. Sometimes these blips are partially
flattened by special purpose hardware but the problem is that
special purpose hardware usually has a short lifespan and is
doomed to financial failure due to lack of economy of scale. The
rest of the time general purpose hardware will catch up. The one
exception I can see here is that the hardware necessary to handle
digital video is special purpose now but will soon be a commodity,
once consumer television goes digital.

Another implication of the Forrest Curve that we're already seeing is
the shrinking, if not outright elimination, of the distinction between
a workstation and a PC. A while ago you could think of a workstation
as a kind of special hardware gizmo that was only bought for a select
few. The rest of us got PCs. But now, with 3Ghz Pentium 4s, the
Infiniband bus, fibre channel disks, and all the rest, it's gotten to point
where the main difference between a workstation and a PC is the size
of the monitor, the lack of IRQs in workstations, and maybe the amount
of memory you can stick in a PC.

The Forrest Curve implies that the folk myth claiming that people's
requirements for computing power expands to consume all available
computer cycles is no longer true. I'm not convinced that it ever
was true, although I have more faith in its corollary about disk
space. Meanwhile, although Moore's Law, which states that the power
of microprocessors doubles every 18 months, seemingly operates
independently, the Forrest Curve does predict that Moore's Law will
start to spread out as the cost of producing ever faster
microprocessors rises.

Another factor I recognize is that having an infinitely fast computer
on your desk doesn't do you any good unless it runs the applications
you need to run. I'm choosing to ignore this issue.

Let's assume you accept the Forrest Curve. What does it mean?

o It means that computer vendors are going to have a tougher
and tougher time selling computers. This is because people above
the curve only need a new computer when something breaks. This
happens less and less often.

o It means that computer purchasing decisions are no longer made based
on price/performance or just performance, like in the dark ages. Now,
when somebody decides to buy a new computer it will be price
alone, or maybe price and service, that determines which computer to
buy. The service aspect should not be ignored. Some people consider
it very important to know that they can call somebody to come to their
home or office to fix stuff and are willing to pay a fair amount for
this. Other people feel that it's important to buy name brands
no matter what the quality of the name brand is. For those of us who
are more enlightened, if our application requires 25 MIPs to run, and
we're trying to decide whether to buy the 30 MIP machine or the 50
MIP machine to run it, the number of people who'd pay much extra for
the 50 MIP machine is very small. Let's face it, nobody is going
to turn down a faster computer but the people making the purchasing
decisions will have a harder and harder time justifying the extra
cost of a faster system for most people. This is especially true in
environments where large numbers of computers are bought for
non-technical people.

o In earlier versions of this document I had the following sentence
right here: "It means that companies like DEC and SGI that are trying
to produce the fastest computer are slowly committing suicide because
there will be fewer and fewer people who need to buy computers this
fast." Two points for me. As of this writing, DEC is gone and SGI is
fighting for its life. Again, in earlier versions of this document
I said here "On the other hand, companies like Sun, and
virtually all PC companies, are doing the right thing by
concentrating on staying just below the Forrest Curve by selling
computers that are fast enough at the lowest price. Although Sun's
approach might have been an accident it has kept them profitable
during some extremely hard times in the industry. Ironically, it may
turn out that Sun will start to suffer too unless it can sell
Sparcstations at PC prices or increase their performance to rise above
the Forrest Curve for a little while". Two points for me again. These
days Sun is hurting and nobody is making money selling PCs except
for Dell, and their reason for success has nothing to do with speed,
little to do with technology, and lots to do with a deep understanding
of logistics and marketing.

So, except for breakage, to be successful the computer industry
is going to have to concentrate on selling to people who don't
currently have a computer. How many people is that in modern
society? Maybe the laptop industry will thrive because it isn't
affected by the fact that so many people already have computers since
most people who buy laptops already have at least one computer.

Maybe computer vendors can postpone hitting the Forrest Curve by
concentrating their marketing and sales efforts into the Second
and Third World but I bet the Forrest Curve still applies there, but
with a lower origin point. Plus, I wonder how much money there is
to be earned there, given hard currency and other non-technical
problems. But, even if these places are exploited, the Forrest
Curve is merely spread out a little. There's simply no way of
escaping it.

Another way to explain the Forrest Curve is as just the commodization
of computer technology. Assuming your application runs on a certain
computer architecture, there's little any vendor can do to add enough
value to get you to buy their system instead of somebody
else's. For all intents and purposes, the different brands of
computers are all the same, just like different brands of flour
and sugar, and buying a computer will be similar to buying baskets
in Tijuana. The only way for computer vendors to survive is to
remember this, and to remember that price and service will be what
makes or breaks them. Caveat Vendor!

Copyright 2002 Jon Forrest. All rights reserved. This document may be
published in any forum for any reason provided the document is not
modified in any way. Last updated (11/7/02).