LinuxSA Mailing list archives

  From: Glen Turner <glen.turner@aarnet.edu.au>
  To  : James Leone <LINUXCPA@netscape.net>
  Date: Mon, 28 Apr 2003 14:50:52 +0930

Re: Steve Ballmer Not Loosing Sleep Over Linux

James Leone wrote:

> <http://news.com.com/2008-1082-998297.html?tag=fd_lede1_hed>

 > SO WHEN IT COMES TO DEVELOPMENT MODELS, YOU'RE CLAIMING THE EDGE?
 >
 > If you want a fix now, we may need to perform better, but you know
 > where to go. There is nobody to turn to if you as a (Linux) customer
 > says, 'I need this.' You can't turn to IBM. They don't write the
 > thing. It's not like IBM can support Linux the way they support the
 > mainframe operating system. They don't write the code for it. All they
 > can say is, 'You can call us and ask us a question, but if you
 > actually want something done we can't do it.'

I can think of at least three people on this mailing list
I could approach with a bucket of cash and a bug fix request
and know the result would end up in the mainline code to
whatever code is at fault.

In comparison I don't know anyone that could get me a code
fix into the Windows code base.

I imagine that's the case for every serious (ie: 50+ boxes)
Linux user.  No Window's user of equivalent size has such
opportunity.  Even bigger Window's users struggle.  For
example Boeing has been pushing Microsoft for years to
get XML support into Word, and look at the pathetic result.

 > WHO IS THE TARGET AUDIENCE FOR WINDOWS 2003 SERVER? IS IT WINDOWS 4.0
 > USERS? THE UNIX CROWD?
 > ...
 > No. 2, there are people moving applications from expensive gear onto
 > cheap gear. I think Windows Server 2003 is going to look good to a lot
 > of that community. Those are high-performance computing applications,
 > or applications running on Solaris or AIX.

Unlikely.  There are roughly 2 Windows clusters and 50 Linux
clusters in the Top 500 supercomputers.  All of the Top10
are UNIX variants, including one Linux cluster.  See
   <http://www.top500.org>;.

Cornell Theory Centre faced significant migration costs
from AIX to Windows, costs which were borne by CTC rather
than the supercomputer's users.  CTC thought the move
worthwhile because of the heavy use of Windows analytical
applications by its Wall Street clients.  Wall St is
pretty unusual in running analytical applications on
Windows rather than UNIX.  However, Wall St is also
pretty unusual in willingness to part with large wads
of cash, something most supercomputer centers find
hard to come by after their initial construction.

Linux clusters are predominately chosen by commercial
supercomputing users, mainly because the biotech and
geological applications are written to use a UNIX
variant and Linux can build a made-to-task supercomputer
cheapest.  These clusters were pretty limited due to
the 32-bit nature of the CPU, and as a result there's
now a deployment of lots of Linux clusters with AMD
Opteron CPUs.

It's unclear if supercomputing clusters made from standard
CPUs are worthwhile in the longer term.  The Earth Simulator,
a cluster of NEC SX-5 vector CPUs, has performance exceeding
the remaining Top10 supercomputers.  And it's yet to be upgraded
to NEC SX-7 CPUs.

With such performance, renting time is on the Earth Simulator
is an attractive alternative to building your own cluster.
This isn't an alternative open those with secret data or to
those running weather applications (as opposed to climate
applications).

Regards,
Glen

-- 
  Glen Turner                (08) 8303 3936 or +61 8 8303 3936
  Australian Academic and Research Network   www.aarnet.edu.au

-- 
LinuxSA WWW: http://www.linuxsa.org.au/ IRC: #linuxsa on irc.freenode.net
To unsubscribe from the LinuxSA list:
  mail linuxsa-request@linuxsa.org.au with "unsubscribe" as the subject