[Leaplist] MarketWatch: The Persistence of Unix

[Phil Barnett]

By Timothy Prickett Morgan

In the physical world, a number of difference forces are at play, and
each in turn displays supremacy, or at least relevance, at subatomic,
atomic, molecular, and wider levels. The weak force allows quarks to
change form, and enables, among other things, nuclear fusion inside the
sun; the strong force binds the nuclei inside atoms together, and the
electromagnetic force glues atoms and molecules together. Gravity is the
weakest of the forces in some terms, but it warps space and tells matter
how to move throughout it. In business, inertia is probably the
strongest force, but so are familiarity, expertise, comfort, and value.
These forces also operate at different levels within an organization,
and they explain the persistence of Unix as a platform two decades after
Unix servers first started appearing in commercial settings.

It is tough to argue that the collective Unix platform reigns supreme in
the data center, but it is safe to say that the remaining Unix platforms
are certainly a preferred platform, if not the preferred platform, for
mission-critical workloads in companies today. Look under the screens of
an ERP system, and the odds are that sitting behind that system is a
Unix box running one of a handful of databases. Unix machines from
Hewlett-Packard, IBM, Sun Microsystems, and Fujitsu-Siemens are also
popular in niche markets. These boxes are not the cheapest on the
market, and with the triple whammy of Windows and Linux operating
systems and very respectable X64 processors, it is reasonable to
question whether or not the three remaining commercial Unix platforms
can persist.

The answer, of course, is that AIX, HP-UX, and Solaris are very likely
to be around for a very, very long time. People were ready to write off
the mainframe 15 years ago, and while the role of the mainframe has been
diminished at IT shops, IBM has been able to keep the mainframe relevant
enough to justify its investments, and more importantly, to rake in lost
profits on hardware, software, and services. For all of its talk about
being platform-agnostic, it is probably safe to say that half or more of
IBM's revenues are derived directly or indirectly from mainframes, and
an even larger portion of its profits come from mainframes.

Good platforms persist, despite immense competitive pressures, and they
do so because they evolve under pressure, or they die. Two decades ago,
just about every company that sold a proprietary server platform jumped
into the Unix racket because it became obvious pretty quickly that open
systems were something that businesses would want. And the
high-performance, low-cost RISC engines in most Unix systems were
another thing that they found attractive.

Unix machines meant having somewhat portable applications and a common
set of programming and system administration skills that translated,
more or less, along with the applications. This allowed nearly direct
competition between vendors, which drove the price/performance curve for
Unix boxes as hard as Moore's Law would allow, and sometimes maybe a
little faster than was sensible given the limits of Moore's Law. All of
the major Unix players from the previous two decades have screwed up
their processor and server roadmaps at one time or another, and some of
those vendors are now dead.

That Unix managed to capture even half of the worldwide server market at
all, considering the alternatives, is somewhat stunning. Various
technical and economic pressures in the server market have made it more
difficult for vendors to get the money they had planned on. Forecasting
fortunes is easy, getting them is not so easy. In July 1999, for
instance, as the Unix server market was roaring during the dot-com boom,
analysts at IDC made some projections about the server market looking
ahead five years. At that time, they projected that the global server
market would grow from $65bn in sales in 1998 to $89bn in 2003, and they
said further that the Unix collective would get $37bn, or about 42% of
the market. Windows, according to this projection, would account for
$27bn in sales, or 30%, and other platforms would comprise the remaining
$25bn.

Of course, nothing of the kind happened. Just like Unix was bolstered by
a worldwide recession in the early 1990s, which spurred the move away
from mainframes and minicomputers, the recession in the early 2000s
pushed companies to rethink their Unix strategies and to instead deploy
cheaper X86 servers running Linux or Windows.

What happened then, of course, would make Unix enthusiasts laugh except
that many of them lost their jobs. Server sprawl ensued, and instead of
having relatively expensive Unix machines that ran at 20% or 30% of
their processing capacity, companies bought lots of cheap X86 and then
X64 iron, which ran at maybe 5% to 10% of capacity. On a
cost-per-unit-of-work basis, there was no improvement at all. And these
days, when power and cooling are such big issues, it is no surprise that
the more sophisticated virtualization technologies embodied in the top
Unix platforms are making companies rethink how they deploy applications
in their data centers.

The fact is, no one wanted to predict the end of the dot-com boom, even
if they could have, because IT vendors were making money hand over fist.
When that bubble burst, there was no way that the server business was
going to grow to $89bn by 2003. According to IDC, worldwide server sales
were $57.5bn in 1995, they grew to $60bn in 1996, peaked at $65bn in
1997, declined  to $63bn in 1998, and declined further in 1999, because
of the ERP-Y2K slowdowns, to around $61.5bn. Sales jumped to an
incredible $67.5bn in 2000 because of the dot-com and telecom booms, and
then crashed by 20% back down to under $50.1bn in 2001, about the same
sales levels as 1994. In the middle of that post-9/11 recession in 2002,
IDC figured that the server market contracted by another 11.6% to
$44.3bn. The economies of the world recovered a bit in 2003, and sales
of server gear rose to $46bn. But by this time, Linux was established as
a cheaper alternative to Unix, with many of the same features and
certainly the same look and feel. Over the past five years, Unix has
been propped up first by IBM's aggressive sales, then by HP's transition
to Itanium, and now by the stabilization of the Solaris platform.

Here in 2007, Unix and Linux together can claim revenue share leadership
over the Windows platform. In the first quarter of 2007, Unix platforms
accounted for $4bn in sales, Linux servers comprised $1.6bn in revenues,
and Windows platforms had $4.8bn in sales. If Linux is Unix, then Unix
is still winning the money market share race. But Unix has been, at
least for now, relegated to specific jobs, such as supporting big
databases or the backbone of financial services, telecommunications, or
other industries where companies are happy to pay a premium for using
the Unix operating system and hardware platforms they know well.

And this, of course, makes perfect business sense. Inertia is not a bad
thing always. It is a kind of energy that allows you to get through the
days, weeks, and months and just get the work done. Change is very
difficult for most organizations, and if they have the skills to make
their current systems work efficiently, then changing to a so-called
"cheaper" platform for the sake of change is idiotic.

By the same token, Unix platforms have to compete directly with Windows
and Linux platforms, and because these are now the volume platforms in
the market, they have to offer more features, more reliability, and as
many applications as they can muster to get those data center dollars.
And so far, the Big Three Unix players have kept their platforms
relevant. IBM has focused on performance and price/performance, and will
be second to none with its Power6 machines, once they are out. Sun has
radically re-engineered its Sparc platform (a work in progress) and
adopted X64 processors to play the price/performance game against other
Unix players as well as against Linux and Windows. Sun has also taken on
Linux directly by going open source with Solaris. And HP finally has
decent performance with dual-core Itanium 9000 processors in its
Integrity line, and has HP-UX v3, which is a high-performance operating
system.

In short, don't count these Unixes out just yet. These three Unixes are
forces in their own right, and they will have to be reckoned with

-- 
Phil Barnett
AI4OF
SKCC #600