[Beowulf] An OT patented rgb editorial rant, skip if you like...

Robert G. Brown rgb at phy.duke.edu
Tue Jan 23 15:23:04 PST 2007

On Tue, 23 Jan 2007, Thomas H Dr Pierce wrote:

> Now as for the size of the computer market and Microsoft. Nope, the top 10
> companies in the US are 1. Exxon Mobil 2. Wal-Mart Stores 3. General
> Motors 4. Chevron 5. Ford Motor 6. ConocoPhillips 7. General Electric 8.
> Citigroup  9. AIG  10. IBM  - and one can argue that IBM is not a computer
> company these days: They are a services company to a large degree.
> Personally I fear Walmart, but the Oil companies are probably more
> powerful.

I agree.  Oil companies are terribly dangerous and powerful, and in
spite of publications to the contrary I strongly suspect that they have
manipulated the last two presidential elections and dictated a
tremendous amount of US foreign policy from behind the scenes at the
White House.

The point isn't that Microsoft is the world's greatest corporate danger,
it is that it is a unique corporate danger, different from those you
list above in many ways.  The biggest single difference is that there is
real competition in all of the markets represented above, indeed in
nearly all major markets including those that are dominated by very
large companies.  In fact, the top ten list above contains a number of
companies that are in fact competitors.  Some of them are indeed
dangerous -- as I said, the world is still finding ways of dealing with
the multinational supercorp.  Still, there are many oil companies (and
potential alternatives to oil, however much we are addicted to the
stuff).  There are many car companies in many countries, and the barrier
isn't so great that one couldn't conceive of a new one emerging.  Some
of these are banks or holding companies, which are definitely dangerous
but are also strongly regulated (not necessarily effectively, but

Microsoft is quite different.  It is basically unopposed in the
marketplace (however much its competitors would like to think otherwise,
the numbers are simply overwhelming), and has agreements in place that
make it nearly impossible for any real competition to arise.  It has
deals with hardware manufacturers (and the practical side effects of its
consumer monopoly) to guarantee that it and only it releases an
operating system product that will run "all PC hardware".  To further
lock this in, it has agreements -- totally legal ones, I would guess --
that lock in the vast, vast majority of computer resellers to offer
exclusively Microsoft Windows as a pre-installed computer operating
system option.  Local vendors that I know of that would LIKE to offer
pre-installed Linux systems cannot do so because if they do, they will
essentially be forced out of business overnight as MS bumps their prices
by more than enough to remove their marginal profitability in this
thin-margin business.  Similarly, desktop software companies that do not
develop products that run on Microsoft's operating system simply have no
chance of surviving.

> And where is Intel in this monopoly?  They own 80% share in the world
> market by some measures. But I do not want to add to the conspiracy

This is a smaller share than Microsoft's, and they have numerous sources
of competition.  Intel dominates CPUs and computer firmware, perhaps,
but they have solid competition there and this is still only one part of
the overall chip market where there is far more competition, including a
great deal of global competition.  Worrisome, perhaps, but consider that
Intel requires raw materials and multibillion dollar chip foundries and
a huge amount of R&D investment to operate in its marketplace.  Chips
have to be built, humans and machines have to build them, they have to
be assembled by humans and machines into devices (usually by an entirely
different company), the devices have to be loaded with ware of one sort
or another, shipped to wholesalers, shipped further to retailers, and
are marked up every step of the way.

In spite of the immense overhead of all of these steps and the immense
investment required to actually build the chips into machines (where
ultimately those chips are commodity items and easily replaced by
functional equivalents subject only to the co-development of firmware
and/or software) what do we find?  A modern PC sells retail for as
little as $500 -- even laptops are now selling for only a bit more.  To
load it with Microsoft Windows XP Pro and Office Pro at full retail
costs MORE than this.  Add antivirus, add any other software at all and
your software costs COULD exceed the cost of the system on which they
run by a factor of two or even more.

This is truly amazing!  The "manufacturing costs" for this software are
on the order of a buck, and more money is probably spent on the box and
manual (such as it is) than on the actual CD(s).  Instead of teams of
hundreds of engineers and billion dollar capital investment foundries
and tens of thousands of employees working in the manufacturing sector
and massive sales and support operations one has teams of hundreds of
(software) engineers, followed by -- sales and support, with as little
of the latter as they can get away with and maintain their market.

They might as well just print money.

So I don't think it is at all fair to compare Intel with Microsoft.  One
has a position of major risk, is constantly required to reinvest huge
blocks of money in ten-billion dollar chunks or lose their market
dominance, and has competition that is doing their best to eat their
lunch.  As a person that screens graduate applications from Chinese
students, let me tell you that Chinese scientific academe from high
school through graduate programs there is overwhelmingly focused on
microelectronics, nanotechnology, and novel information processing
schema.  Intel, AMD, TI, Motorola, Fujitsu, IBM -- none of them are
secure, not from each other and not from new threats being born in the
world marketplace.  In this competition -- even competition between
giants -- world consumers gain tremendous benefits.  That's WHY the
margins on computer systems are so thin -- it is one of the most
cut-throat businesses on the planet.

But Microsoft doesn't care.  Every one of these marginally profitable
machines that is sold is a guaranteed $50, $100, $300 in their pocket
over the lifetime of the machine, the bulk of it pure profit.  They make
more actual post-cost money from the sale of a computer than any other
participant in the process.

> theories. A source of conspiracy theories is when people see short-term
> tactical events that personally affect them that are driven by overlooked
> long-term trends.  Microsoft is benefiting from the trend in personal
> computing. This trend could end with computing becoming entertainment ala
> youtube, or personal cellphones or very powerful personal digital
> assistants or personal networks or something else. With the rate of change
> these days is it unlikely to remain a trend in personal computing.

These examples seem to me to be fairly irrelevant.  They are all
distinct markets and have nothing to do with the viability or necessity
of personal computers in business, nor with the software packages that
will support those desktop business functions.  Microsoft is uniquely
positioned to exploit their market position and maintain dominance in
any new or emerging area as it emerges, as well.

However, the general idea of an emergent challenger is certainly
something to hope for.  However, note well that Microsoft has endured
more or less unchallenged since it betrayed IBM on the OS/2 deal in the
early 90's, and has in that time wiped out OS/2, Netscape (co-opting the
consumer/client side of the web), numerous other software companies,
concepts, products, and survived even their own incompetence -- who else
could sell an operating system that you don't dare to use in a
professional environment without spending money with third parties to
"fix" its huge, glaring security holes? Fifteen years is three computing
eternities already, and it is difficult offhand to see them failing in
the next two eternities UNLESS the long awaited invasion of the penguin
people occurs.

It could -- IBM is pushing it (amazingly clumsily, in my opinion, given
their capacity for core investment).  Novell could possibly manage it,
although they have a history of shooting themselves in the foot.  Red
Hat is pursuing the most conservative of strategies and trying to become
Sun Microsystems, not Microsoft, which is worrisome as they are
duplicating a failed strategy, badly.  Linux "competition" for Microsoft
these days reminds me unfortunately of the three stooges, not of a clear
and coherent view of the challenges involved in achieving world
domination.  However, linux doesn't stand or fall with its corporate
champions -- it is too delocalized, too robust, too free.  And at any
moment, any of its corporate champions could grow a brain, or
demonstrate that I'm wrong and they've had a brain all along and it is I
that is brainless or lacking in subtlety.

> As for Microsoft HPC, well, that could work. I still remember the days
> when everyone said that white collar workers won't type their own memos,
> when people said that no one would buy personal computers because
> operating systems were complex and when people said that home networks
> were too complex to setup and use. ...  Microsoft HPCs have barriers to
> overcome, but if they work for a segment of the poplulation, then let's
> move on to the next issue.

Oh, I expect them to be wildly successful, but that isn't where the real
money is.  Rather they are planting a sprout, and doubtless they will
water it well, so that however well it grows or productive it proves, it
will stunt the growth of everything else in the garden.  Except that
they don't really >>understand<< linux, or the forces that gave rise to
COTS clusters in the first place.  Their cluster effort remains
vulnerable to the same things that old big iron supercomputers were
vulnerable to -- long term price pressure.  Again, Microsoft is counting
on people being willing to spend a signficant fraction of the cost of
their hardware on their software instead.  This means that they have the
eternal choice -- more boxes (and run linux) or fewer boxes (and run
Windows).  At constant investment, they will be spending long term
throughput capacity for the software.

Some groups will be happy enough with this deal.  Specifically, those
who want intermittant "bursts" of HPC and who also have a MS-dominated
operational environment.  However, groups who expect an uptime of close
to 100% on the resource will not be happy losing 20% of their potential
throughput (or more) just so that they can submit through a pretty GUI,
even for smallish clusters, and face it -- building a Linux cluster just
isn't that difficult.  High school students do it successfully.  I've
helped maybe fifty groups around the world get started just with a few
email messages and some guidance, and then there is this list.  And it
is pretty easy to get a "pretty" prebuilt turnkey linux cluster as well,
and the margins are likely to be smaller relative to the hardware.

Basically, as long as MS continues to dominate the desktop market,
they'll sell HPC systems to MS-centric clients.  If their grip on the
desktop market should loosen, or if their share of the server market
starts to slip (which basically can only happen as organizations adopt
linux and lower the management barrier cost for doing linux clusters),
their cluster market will slip along with it.

Or maybe this is just wishful thinking on my part.  Punditry isn't
perfect, and the world is a complex place...;-)


> ------
> Sincerely,
>   Tom Pierce

Robert G. Brown	                       http://www.phy.duke.edu/~rgb/
Duke University Dept. of Physics, Box 90305
Durham, N.C. 27708-0305
Phone: 1-919-660-2567  Fax: 919-660-2525     email:rgb at phy.duke.edu

More information about the Beowulf mailing list