Beowulf Training....

Robert G. Brown rgb at phy.duke.edu
Tue Jan 7 12:38:12 PST 2003


On Tue, 7 Jan 2003, RANGI, JAI wrote:

> HI
> 
> I am looking for some sort of training on Beowulf cluster and parallel
> programming. Also I want to know the areas where Beowulf can be used in
> Educational institutes in term of different programs as well as
> research. 

Wake Technical Institute outside of Raleigh, NC has a brand new program
in this.  They are partnered and co-recipients of an NSF grant with Maui
Community College in Hawaii, Pellissippi State Technical Community
College in Tennessee, and Contra Costa College in California.  Each of
these small schools is partnered with a supercomputer center or national
lab -- NCSC, Maui HPCC, Oak Ridge, Lawrence Berkley respectively.  I
informally consulted on their proposal and thus am fortunate enough to
know about it.  I believe that all of these schools are putting together
two year (?) programs to cover cluster design, parallel programming, and
cluster support.  The one at Wake Tech has a variety of RTP and Red Hat
persons on their advisory board (as well as, de facto, me:-).

At a higher level, Clemson has an excellent cluster program with a
number of respected faculty on this list.  Duke has a program that has
been around for a while working on exotic topics but is currently moving
a little more mainstream.  There are probably other programs out there
at the four-year university or graduate level -- maybe their owners will
speak up.  LOTS of universities have clusters now and hence have
upwardly mobile pathways for on the job training if you are skilled in
basic systems admin.

Scyld also offers training, I believe, although I don't know how broad
it is or whether it leads to any sort of certification.

Parallel programming per se I'm not as familiar with, although I
probably should be.  I'd expect schools "near" a supercomputing center
to all have at least a class or two.  Probably somebody else on the list
knows.

The answer to your second question is too large to even summarize.  Ten
years ago I was using our departmental SunOS/Irix LAN as a cluster
supercomputer via PVM and scripts.  Seven years ago Duke had pretty much
one "beowulfish" linux cluster -- mine -- and a CPS cluster that wasn't
devoted to "supercomputing" per se.  Today the administration is having
a hard time counting the number of clusters in operation on campus, and
is openly encouraging more with lots of interested parties coming to any
sort of meeting where they are discussed.  Clusters are being used to do
physics, mathematics, quantum chemistry, various kinds of biology,
statistical simulations of all sorts -- and that's just here.  A better
question would be -- where within the sciences, mathematics, statistics,
and even some of the humanities are clusters NOT being created whereever
a compute need that exceeds the capacity of a single CPU exists?  Not a
lot, it seems.

There is indeed a gap between this growth in application and the human
resources needed to support it.  Would be cluster-builders also tend to
underestimate or ignore the hidden costs (humans, power, AC, other
infrastructure) or view them as opportunity-cost advantages rather than
real expenses.  This is one of many things Duke is addressing at a high
level -- how can Duke encourage all the advanced science clusters can
facilitate while controlling costs and not exploiting already
overburdened LAN sysadmins?  What is the "cost-benefit" of all the
various clusters to the University as a whole in terms of e.g. grant
overhead, etc. (noting that the answer COULD turn out to be anywhere
from greatly positive to somewhat negative -- not obvious, at any rate)?

   rgb

> Any information will be appreciated.
> Thanks
> 
> Jai Rangi
> ------------------------------------------------------- 
> In the world with no fences, why would you need Gates ? 
>                                           - Linux 
> ------------------------------------------------------- 
>  
> 
> _______________________________________________
> Beowulf mailing list, Beowulf at beowulf.org
> To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
> 

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






More information about the Beowulf mailing list