[Beowulf] about clusters in high schools

Robert G. Brown rgb at phy.duke.edu
Fri Jan 27 10:06:22 PST 2006


On Thu, 26 Jan 2006, Brian D. Ropers-Huilman wrote:

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: RIPEMD160
>
> My first cut response, not the RGB 'bot response, which I'm sure will be
> full of excellent anecdotes, is: absolutely.

:-)

Absolutely indeed.

My "excellent anecdotes" on this subject are basically derived from:

   a) The experience of guiding maybe a couple dozen high school kids at
various points in time through the building of their first beowulfs,
usally with little or no support from their schools, all done offline.
I get a lot of people who just write me directly saying 'hey, I want to
build a beowulf, how should I go about doing it'.

   b) Experience derived from directly supporting a community college
effort to build a beowulf engineer/administrator training program (where
Wake Tech CC outside of Raleigh is the local branch of this NSF funded
initiative).  I have met several high school instructors who were
interested in self-training themselves to where they COULD support such
a program at the high school level at various meetings and symposia
presented within the program.  This has taught me a very serious other
side of the story.

   c) My knowledge of what most schools have and can provide in terms of
infrastructure at least around here in NC.  Not too easy to extrapolate
to "all schools" but again, it gives me some insight on the difficulty
of establishing suitable programs.

>From this, there is good news and bad news.

The good news is that bright kids DO like to build beowulves in high
school (including in schools in e.g. India, not just in the US!).  In
nearly any school you'd have 5-15 students who would be perfectly happy
to immerse themselves in it and have a great time doing so with ANYTHING
like encouragement.

The bad news is that the ones who succeed generally do so without any
meaningful support from their school.  Sometimes not even with access to
school-owned machines as a resource.  Almost never with anything like
mentorship within the school itself.  They scrounge machines themselves.
They find switches.  They learn about linux (usually from me telling
them EXACTLY how to install a functional version for free on their
scrounged hardware).  They find toy problems to play with.  Then alas,
they graduate and move on, leaving very little that survives or might be
used to turn into a "program".

Why so bleak a picture?

Well, for one thing Windows overwhelmingly dominates as the OS installed
in most schools.  It is so pernicious a phenomenon that they don't teach
"spreadsheets", they teach "using Excel".  They don't qualify students
with an end of grade test on "word processing", they qualify students
with a test on "using Microsoft word".  That this is Evil beyond all
measure is beyond any doubt -- imagine the screams if one had to take
all drivers tests in a state using a Ford.  On the other hand, the
schools are crippled by the near-vacuum in computer competent teachers
in general -- it is doing as much as they can to end up with somebody
that can teach "using Word" or "using Excel" as part of "keyboarding".

Note that few schools even have a local systems administrator.  They
hire out all of the management of the school's networks to an outside
contractor, who locks down everything and is responsible for securing
everything and is MOST NEGATIVE about the thought of kids building a
supercomputer system "inside" the school's network where it could wreak
untold mischief.  A point of view I'm not totally negative toward, by
the way.

Even the relatively progressive schools that DO know what linux is, that
DO have a faculty person with some experience in linux (rarely
"professional grade" experience, unless the person involved is a true
saint, as anybody with pro grade experience can make 3-4x the salary of
a high school teacher without even trying hard), don't have the LEVEL of
experience in networking and supercomputing to be able to support a
beowulf program meaningfully.  That's what the Wake Tech thing showed
me.  Teachers from schools would need e.g. community colleges LIKE WTCC
with programs where they could teach the teachers, before there will
ever BE any teachers that can teach this as a course or part of a
meaningful high school experience with some continuity.

The long term solution to the problem, perhaps, is to do what Doug and I
and many others have been working on for years -- create a sufficiently
robust and strong set of web based resources, including the PEOPLE (many
of whom are on this list) who are able and willing to act as mentors, as
teachers of teachers, as supporters of CC programs to formally train
teachers -- that one can bootstrap the process, where any bright student
CAN build a beowulf at a school, where a faculty person working with
them CAN learn about linux and supercomputing in the process, where
there IS a chance that a program can be born out of the experience.

Fortunately, building a simple beowulf is pretty easy -- it CAN be done
"almost" from a recipe.  Things like warewulf make it easier than ever,
almost to the point where one can say "boot all these boxes with this
CD, starting with the server node, answering questions as the nodes come
up" to build a cluster with "no prior knowledge" of linux or cluster
computing.  That shrink-wrapped level is where it needs to be to get at
least SOME groups started.  From there there are plenty of pathways to
learning what's actually happening "under the hood", ways to learn about
process parallelism, task scaling, system administration, networking and
much more.  It IS a tremendously rich subject that primes a student for
almost any kind of career or college program in IT-related studies in
science, social science, math, or even the traditional humanities.

I keep hoping that we see more of this, but I also recognize that the
obstacles are still pretty significant.  Ultimately it may be the
parents, or changing the way the Government recognizes "standards", that
drives a movement to linux in schools and the consequently greater
degree of learning about systems that would enable.  I see that
happening in Europe and Asia, but not (alas) so much here...

     rgb

>
> My sources say -- pay attention to the upcoming State of the Union address.
> You'll hear a lot of talk about U.S. competitiveness in STEM (Science,
> Technology, Engineering and Mathematics). HPC is used more and more
> regularly in "traditional" businesses such as finance or even package delivery.
>
> Modeling or simulation is becoming the third pillar of Science, next to
> theory and research [ Ref.: pg. 12 of the 2005 U.S. PITAC report
> "Computational Science: Ensuring America?s Competitiveness" ].
>
> While this whole "competitiveness" bit smells strongly of "the sky is
> falling," I happen to believe strongly that, regardless of our weakening or
> strengthening of STEM graduates in the States, HPC is steadily moving it's
> way out of academe and into every day business.
>
> The Council on Competitiveness did an excellent short video, using the
> Penguins from the movie "Madagascar," on "HPC in Everyday Life" [
> http://www.compete.org/hpc/everyday_life_video.asp ]. {full disclosure: we
> contributed to this video}. For $10, it's a cheap way to possibly further
> influence the School system or wary parents.
>
> H.Vidal, Jr. said the following on 2006-01-26 21:25:
>> Howdy.
>>
>> My son attends a Science and Tech focused high school here in beautiful
>> New Jersey. This is a pretty neat place for a high school, about 70%
>> of the faculty has their PhD Kids take about 2-4 semesters of physics
>> and chemistry, there are lots of computers, they teach Scheme as well
>> as C++, Java, etc. Freshmen get the option of taking things like Number
>> Theory. Interesting place.
>>
>> However, I have a thought. There is, to my knowledge, essentially
>> zero exposure to high-performance computing at this school. And I
>> think this is a mistake.
>>
>> My thinking is this. I have observed that in materials science,
>> in medical imaging, in genetics, even in theoretical mathematical
>> studies, these days you see a lot of applied high-performance
>> computing. I get the impression (back me up here if it's otherwise)
>> that skills in high-performance computing have a fair amount
>> of value, and are growing in terms of overall industry demand.
>>
>> Yet smart kids really have very little exposure to these classes of
>> problems, even if there are exposed to the problems themselves.
>> These kids can take a class in genomics, and they even learn about
>> some classes of problems in genomics or proteomics where you
>> need to run large mathematical problems to get 'concrete results'
>> towards practical studies or applications in the problem domain, but
>> they are kept far from actual hands-on or low (or even high)
>> level theory in terms of actual implementations or even
>> engineering considerations WRT HPC.
>>
>> Yet they have *rooms* full of computers doing nothing, fully
>> networked. (there's always lots of rooms of unused computers
>> in places like these, I have found, because they basically keep
>> upgrading to new hardware every year or two. Each summer,
>> the hallways are nearly impassable due to stacks and stacks
>> (not kidding) of computers to be thrown out or recycled).
>>
>> So I have convinced the faculty at this school that HPC
>> is enough of a valuable study, even a strategic interest, that
>> sharp kids like these really should be educated in the ins and outs
>> of high performance computing. In general, HPC; in particular, our
>> good friends, the Beowulf clusters.
>>
>> I would like to get real feedback from students, engineers and
>> scientists on this list about this broad idea: is it useful to expose
>> young engineer and scientists-to-be to HPC at the high school
>> level, in generaly, but especially if these kids are on track
>> to be the next generation of users of this tech? If so, what is a decent
>> route to take when it comes to this as a HS level scholastic pursuit?
>>
>> So there you go, I have thrown out the first chip. Any takers to place
>> a comment or two?
>>
>> Thanks in advance for your collective wisdom and help.
>>
>> H. Vidal, Jr.
>> _______________________________________________
>> Beowulf mailing list, Beowulf at beowulf.org
>> To change your subscription (digest mode or unsubscribe) visit
>> http://www.beowulf.org/mailman/listinfo/beowulf
>>
>
> - --
> Brian D. Ropers-Huilman, Director     .:.    High Performance Computing
> Networking, Infrastructure, & Research / Information Technology Support
> Frey Computing Services, Rm. 323                        bropers at lsu.edu
> Louisiana State University                          +1 225.578.3272 (V)
> Baton Rouge, LA 70803-1900  USA                     +1 225.578.6400 (F)
> http://www.hpc.lsu.edu/                         http://www.cct.lsu.edu/
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.2 (GNU/Linux)
>
> iD8DBQFD2Y7QwRr6eFHB5lgRAwh/AJ4gqfcDUNHRsR2ZrL83HUuSRe7CJQCfcd7J
> wICEye1RWD3eGvORmBAtzKI=
> =teYU
> -----END PGP SIGNATURE-----
> _______________________________________________
> 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