WinNT/Linux hybrids?

[Schilling, Richard]

Hey, Peter Dell should chime in on this one.  This really edges on getting
Beowulf accepted into the business mainstream.  

<STRATEGY ISSUES>

Robert is 1020% correct about latex, and not having to utilize Office
products, etc . . . for science types, it's optimal.  And Dr. Knuth gets a
lot of credit for developing tex too.

A good example of integrating solid clusters with business would be our next
system.  It uses a cluster built on DataGeneral servers, and will be
installed within 18 months.  A requirement of the product is that it
delivers data to the desktop, which means Office product integration or
Windows desktop delivery using client/server.  Those types of systems are
excellent case studies.  The system also demonstrates that while Microsoft
has market share on the desktop, it dosen't hold a candle to midrange
clustering, mainframe clustering, and Beowulf clustering as well.

What I have run into, and this is a big issue for business
heads/administrators/etc . . . is that users don't care to stretch too much
technically when it comes to getting work done.  I've had to deal with that
in various ways, the most notable is integrating and Excel spreadsheet
directly into a relational database -yech.  Users just want to get that
report out/do that analysis so the boss does not harp on them too hard.  On
the other hand, they are usually open to something that is easy, dosen't
cost money, and gets the job done better.  The key is to know where the user
is currently, and deliver a product on a silver platter, all without
degrading the performance of the cluster.

If we conquer the Office products, and integrate the Beowulf power so the
average person can "reach out" and grab it, then it'll be much easier to
sell to the corporate types.  This could be as simple as getting Excel to
read Beowulf results off of a node through Apache, or writing an export
function to port Access databases to a distributed database.  It could be as
complicated as porting PVM to work on Windows 95 machines without requiring
remote logins.

Just my $0.02.

</STRATEGY ISSUES>


Richard Schilling
Webmaster / Web Integration Programmer
http://www.affiliatedhealth.org
Affiliated Health Services
Mount Vernon, WA USA
360 856 7129



> -----Original Message-----
> From: Robert G. Brown [mailto:rgb at phy.duke.edu]
> Sent: Monday, February 26, 2001 12:01 PM
> To: Jim Lux
> Cc: Beowulf (E-mail)
> Subject: Re: WinNT/Linux hybrids?
> 
> 
> On Mon, 26 Feb 2001, Jim Lux wrote:
> 
> > What would be the possibility of building a cluster using 
> WinNT workstation
> > (or NT server) on the head/worldly node and Linux on the 
> (diskless) cluster
> > nodes.  You'd use one of the many NT ssh or Xwindows 
> servers to talk to the
> > nodes.  Is it possible to netboot Linux from a NT server 
> (what has to be
> > running on NT to make this work)?
> >
> > Why would one want to do this?  There is a huge installed base for
> > powerpoint, excel and word, and eventually you have to get 
> your  results
> > into that form.  Also, you could start your cluster 
> grinding on a lengthy
> > task, effectively in the background, and continue using all the
> > "institutionally supported" software (i.e. Office, Outlook, 
> Project, etc.)
> > in the foreground.
> >
> > Sure, it's far from optimal, but then, what separates 
> theory from practice
> > is that practice has all these warts and impracticalities 
> but it does useful
> > work.
> 
> <constructive comment>
> It would be far, far simpler to just export a "results" 
> directory to an
> NT workstation on an outside network than to run NT on the head node
> native and hybridize the cluster.
> 
> It would also be far cheaper -- human time is the most 
> expensive single
> resource and it would take a lot of it to make a true NT head 
> node work
> transparently with linux slave nodes without flakes.  
> Debugging problems
> across two operating systems and subroutine library sets and 
> TCP stacks
> is not something I'd undertake lightly.
> </constructive comment>
> 
> <rant>
> 
> Finally, one does NOT have to get one's results into 
> powerpoint, excel,
> or word form.  I've managed for years to make do without any of the
> three and published and presented all sorts of scientific work in the
> meantime.  After all, Physical Review accepts latex submissions, for
> good reason.
> 
> Itemizing, spreadsheets are a dime a dozen -- linux has some very good
> ones.  I personally detest word processors in general for 
> technical work
> -- latex is far better -- and I REALLY hate Word.  WYSIWYG document
> processors unleash an author's creativity to the point where they
> produce terribly formatted documents because they didn't pay attention
> back in 8th grade where they were supposed to learn all the rules for
> correctly formatting documents.  Latex is designed so that you don't
> HAVE to know how to format a document and you'll get correct and
> consistent formatting every time.  You do have to learn latex, of
> course...;-) Power Point shares some of the problems of word, some of
> the problems of Excel -- if it is adequate for what you want 
> to do with
> it, fine, but I find that latex is also superior for laying out talks
> and transparencies unless one is in marketing and sales and need cute
> little colored triangles jotting across one's text.
> 
> </rant>
> 
>    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 at phy.duke.edu
> 
> 
> 
> 
> _______________________________________________
> Beowulf mailing list, Beowulf at beowulf.org
> To change your subscription (digest mode or unsubscribe) 
> visit http://www.beowulf.org/mailman/listinfo/beowulf
>