Accounting for number of Linux & NT Clusters in the world

[Robert G. Brown]

On 20 Oct 1999, A.J. Rossini wrote:

> 
> 
> 	<-- many of Robert's fine points deleted -->
> 
> >>>>> "RGB" == Robert G Brown <rgb@phy.duke.edu> writes:
> 
>     RGB> On Tue, 19 Oct 1999, Tony Skjellum wrote:
> 
>     RGB> The (literally) bottom-line point is: Let us do a cursory
>     RGB> comparison of the cost/benefit of an NT-based cluster and a
>     RGB> linux-based cluster (beowulf or not), presuming that both
>     RGB> need:
> 
> 	<-- another fine list deleted -->
> 
> Anyway, one point for cost/benefit that's being ignored is human
> time.  lack of unix experience can be a killer, esp if your primary
> admin people are NT based.  Humans aren't really interchangeable.

I agree, except for a very few things.  One is that human costs are more
or less fixed -- linux people or NT people you need pretty close to the
same number of people.  Second is that humans, guided by an intelligent
vision, are perfectly capable of being cost-effectively retrained into
linux experts without loss of service.  Linux can be easily enough
learned by completely untrained high school students, let alone
presumably trained IT professionals who already hopefully understand a
lot of what is going on.

Taking the two things together, it is STILL crazy (in most cases) not to
invest the time and resources required to at least nucleate a possible
conversion to linux in most larger organizations, no matter how well
entrenched NT is in the corporate culture or IT admin group.  After all,
the annual cost savings are immense and forever; the cost of the human
time required for the conversion is finite and a one time expense,
PROVIDED ONLY that no loss of functionality or service occurs (changing
the benefit side of the cost/benefit equation).

This much is pretty easy to see.  So why doesn't the change occur
faster?  Because one shouldn't make the mistake of thinking that the
thing being optimized by any IT organization is either costs (to the
organization) or benefits (to the organization's users of IT systems and
services). Far from it.

The thing being optimized and protected is a superorganism -- a
collection of individuals who appear on the service to be behaving
rationally and will argue that they are doing so, but in fact are
defending the integrity of the superorganism of which they are a part.
In this case, the superorganism making the decision is an NT-based IT
group.  

This entity (regardless of the individual feelings of particular
individuals within or outside pressures or realities) correctly views
conversion to linux as a risk to its existence.  It is totally
unconcerned about overall economies as its budget (usually received from
"outside" this superorganism) is representative of the power and
importance of the superorganism in the larger superorganism of which it
is but a part.  If it uses less money, its power paradoxically goes
>>down<< not up.

Similarly, the idea that conversion to linux requires retraining of its
individual members is intolerable.  During this retraining period the
organism is vulnerable and cannot protect its members.  It is dependent
on the trainers.  Masters are reduced to students and lose face and
privilege.  Once again, the power of the superorganism is diminished,
especially the power and control exerted by the superorganism's oldest
and most powerful individual members, who usually are the one's most
immersed in that superorganism's meme, or raison d'etre and who have the
hardest time converting to a new one with their status and purpose and
relative expertise intact.

Finally, the idea that linux might actually be >>more<< powerful than
NT, and might require >>less<< management (because, for example, it
actually doesn't crash constantly and because serious problems get fixed
with the help of a much larger, faster, and more dynamic superorganism
like the various linux lists rather than being sent to certain death in
the Microsoft amoeba) is absolute anathema to the IT superorganism.
Even more than budgets, power in corporate entities is measured by the
number of people controlled.  Never in the history of corporate
management has a section administrator said to his/her superiors "Gee, I
think we should install product X which does the work of my entire task
force except for maybe one young guy so the money you are spending on
the rest of us can be saved and I can be laid off."  Even if that person
DID say that to their superior, the superior would reject the proposal
until there was literally no choice due to pressures placed by his/her
competitors.  Loss of people (however economically advantageous to the
company, the research group, the university) is equivalent to loss of
power and control.

This has been seen time and again in the IT world over the last two
decades at least.  First it was mainframes vs minis.  Then it was
mainframes and minis vs PC's and later unix workstations.  It was Cobol
vs C, C vs C++, C vs Fortran, all of them vs perl, perl vs java, DOS vs
Macs, Macs vs Windows, Windows vs Unix, NT vs OS2, NT vs Unix, and now
NT vs linux.  In each and every case (and the many subcases ignored) an
entrenched superorganism was met with an economically more efficient
competitor and responded with a class hostile and protective rejection.
In each and every case where the conflicts have been resolved (many are
still pending and some are not direct conflicts and may end up
coexisting in rough equilibrium) the reactionaries are eventually swept
out of power >>provided<< that there are real advantages to the proposed
replacement.  Evolution in action, of course, but the evolution and
competition and natural selection is of superorganisms and paradigms,
not of individuals.

So I don't take the idea that there are really good reasons to run NT in
the long run too seriously.  In the short run, there may well be
benefits that are worth the costs (such as a critical application that
really just doesn't exist for linux) but there is much more likely to be
hostile rejection to defend an NT-based superorganism's perceived
integrity.  History shows us that these efforts are doomed to failure
UNLESS NT develops and maintains some true survival advantage -- a true,
user benefit -- relative to linux.

Ain't happenin'.

    rgb

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@phy.duke.edu



-------------------------------------------------------------------
To unsubscribe send a message body containing "unsubscribe"
to beowulf-request@beowulf.org