[Beowulf] Re: Beowulf Digest, Vol 50, Issue 12

[Mark Kosmowski]

>  Message: 2
>  Date: Fri, 04 Apr 2008 15:14:56 +0200
>  From: Toon Moene <toon at moene.indiv.nluug.nl>
>  Subject: Re: [Beowulf] How Can Microsoft's HPC Server Succeed?
>  To: Jon Forrest <jlforrest at berkeley.edu>
>  Cc: Beowulf Mailing List <beowulf at beowulf.org>
>  Message-ID: <47F629D0.4090807 at moene.indiv.nluug.nl>
>  Content-Type: text/plain; charset=ISO-8859-1; format=flowed
>
>  Jon Forrest wrote:
>
>  > But, the question remains. How can Microsoft compete with free?
>  > How much better will they have to be than standard Linux
>  > clusters before they get any mainstream interest? What technical
>  > features could they add that couldn't be added to a Linux
>  > cluster?
>
>  By becoming as good as Intel's Fortran compiler (when compared to
>  gfortran/g95) ?

I don't know that this is a perfect example, as until very recently,
the GNU Fortran did not support Cray style pointer extensions, which
is a go / no go feature for my clustered application.

>
>  Granted, ifort is free for non-commercial work, but all others pay cash.
>
>  Yet I have to see one instance where the ifort development group set up
>  "gut the facts" campaigns or tries to buy governments to not use
>  gfortran or g95.
>
>  I know the head of the group personally (via the Fortran Standardization
>  Committee J3).  He asks intelligent questions, like "my people are told
>  that gfortran supports feature X of Fortran 2003, while we don't - is
>  that true", instead of bamboozling me with misinformation, fear,
>  uncertainty and doubt.
>
>  Who knows, perhaps one day Microsoft will learn that to be in the
>  software business, instead of being an abusive monopolist, means to
>  write good software.

You're joking, right? (This is itself a joke.)

It might be ironically humorous to see whether this proposed product
fails due to the very thing MS counts on to maintain Windows market
share on the desktop segment - user familiarity.

I mentioned above that until fairly recently, the GNU Fortran was not
an option for me.  The only reason that I tried using gfortran once it
became able to support Cray pointers was that I was too broke to buy
an upgrade license to the PGI compilers and was trying to eek a little
speed out of my program so I started looking again at the freely
available compilers.

If PGI had a lower upgrade cost, I would have likely shelled out the
cash and stayed with them, as I am not a developer and had already
learned what I needed to learn to compile my code with the PGI
compiler.

At this point in the game, it seems that existing clusters are being
dominated by Linux.  MS seems to carry the desktop market by "The
other choices aren't Windows - you'll need lots of training cost to
learn the not-Windows OS."  Now it is trying to enter the cluster
market, where it is facing two strikes out of the box - not free and
not Linux (that most cluster admins are experienced with and will have
a large training cost to learn the Windows cluster OS).

>
>  --
>  Toon Moene - e-mail: toon at moene.indiv.nluug.nl - phone: +31 346 214290
>  Saturnushof 14, 3738 XG  Maartensdijk, The Netherlands
>  At home: http://moene.indiv.nluug.nl/~toon/
>  Progress of GNU Fortran: http://gcc.gnu.org/ml/gcc/2008-01/msg00009.html