[Beowulf] How Can Microsoft's HPC Server Succeed?

[Chris Dagdigian]

I still think MS HPC has a shot in this space.

Remember the size of our market is pretty small -- a few thousand  
installs would have a pretty significant impact in this space although  
within MS that may qualify as an embarrassing product failure.

To me the market for MS HPC products is with commercial ISVs that sell  
software but are currently buried under infinite permutations of  
hardware and software they need to support. Setting up a cluster,  
managing the various communication stacks, application integration and  
batch scheduling is still quite a bit of work.

With a MS shrink-wrap HPC product there is the possibility that an ISV  
can offload much of that responsibility to Microsoft - or more  
reasonably some other software/support/consulting shop that has the  
necessary domain experience to handle end-user application integration  
and usage questions. Right now for commercial ISVs the support  
question is non trivial and involves many different sets of compilers,  
MPI implementations, transport fabrics and cluster schedulers.

Anyway, the possibility that commercial vendors of engineering/ 
scientific software could support MS HPC as a reasonable base that  
allows for a single point of contact (or finger pointing) for  
integration/usage issues is really the only thing I can come up with.

In the market I'm familiar with (life science) there are all sorts of  
devices that could benefit from a shrink-wrapped compute solution.  
Even the smallest labs can get grants to purchase the latest $500K  
confocal microscope or next-generation DNA Sequencing box. Each of  
those devices can spew out a terabyte per day of raw data and many  
times that stuff needs to be post processed and distilled down into  
different forms. A nice little 8-core box running a shrink-wrap HPC  
product with a single support contact could find a nice little niche  
in non-datacenter areas where significant compute is needed nearby  
some other sort of dedicated instrument or device.

Personally I'll stick with Unix and Grid Engine / LSF myself but I do  
see some areas, use-cases and markets where MS HPC could benefit. The  
window of opportunity is shrinking though.

My $.02 of course.

-Chris




On Apr 2, 2008, at 6:04 PM, Jon Forrest wrote:
> First of all, I like Microsoft, and I voluntarily use
> Vista as my desktop of choice. I've built and run the
> Windows environments for the top CS and Civil Engineering
> departments in the US, and I was the first to port
> Postgres to Windows NT.
>
> That said, I just don't see how Microsoft's HPC server
> can succeed. I'm not saying this for technical reasons,
> as I'm sure Microsoft, with enough work, can build
> a clustering environment that will work just fine.
> But, why would anybody buy a Windows cluster when
> there are so many great clustering environments (e.g.
> Rocks, Perceus, Unicluster Express, ...) and so many
> cluster-related packages (*MPI, SGE, PBS, gcc,
> Torque, ...) available for free? What's more, from
> what I can see, there is very little non-Microsoft-sponsored
> development going on in HPC computing.
>
> Microsoft recently announced (somewhere, I can't find it)
> the availability of a test cluster for universities to
> use for financial applications. I bet they get some
> interest since many business schools use Windows, plus
> the cluster is free.
>
> 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?
>
> Cordially,
> -- 
> Jon Forrest
> Research Computing Support
> College of Chemistry
> 173 Tan Hall
> University of California Berkeley
> Berkeley, CA
> 94720-1460
> 510-643-1032
> jlforrest at berkeley.edu
> _______________________________________________
> Beowulf mailing list, Beowulf at beowulf.org
> To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf