[Beowulf] why we need cheap, open learning clusters
Lux, Jim (337C)
james.p.lux at jpl.nasa.gov
Sun May 12 16:27:52 PDT 2013
On 5/12/13 1:59 PM, "Mark Hahn" <hahn at mcmaster.ca> wrote:
>on the original topic, "cheap open learning clusters" to me sounds like
>a great application for EC2 or similar IaaS.
Not really.. There's that intangible benefit from having the hardware in
front of you to do with what you will.
And, more importantly, you pay by the minute for "cloud" services, so in
an educational environment, there will be pressure to allocate a certain
amount of computing services for each student to use.
At UCLA in the 1977-79 time frame, they justified the strategy of "you get
N runs per quarter" as "forcing you to do good work, so each run counts;
careful desk checking before you run, etc.". In reality (so I understand,
not from first hand knowledge) it had more to do with the "computing
budget" assigned to the class.
At UCSD in the 1976 time frame, they said "go in the lab with the LSI-11s
and do whatever you want whenever you want, just make sure you turn the
assignment in on time" (you could also go run your job on the big
Burroughs B6700, a fascinating machine, and there was no limit on the
number of runs or time.. I spent a lot of time trying to figure out ECAP
in FORTRAN using a student account)
Maybe that's because of the difference in which departments things were
At UCLA was Math/CS, at least as an undergrad ("all a good mathematician
needs in the way of tools is paper and pencil"), (notwithstanding that the
intro courses were called Engineering 10,20,30 and held in Boelter Hall,
and there was an actual graduate CS department)
At UCSD it was Applied Physics and Information Sciences (formerly
Aerospace and Mechanical Engineering Scienes - AMES) - definitely a
department where hands on the hardware is important.
Google is my friend today, and I found some interesting stuff:
"Bowles wanted numerous machines to give students hands-on experience
with interactive computing. The university instead committed to the
purchase of a large IBM mainframe and set the center's priorities on
business process support for the university administration"
Ken Bowles is the guy responsible for UCSD Pascal, and now I know why he
pushed for it.
I suppose, indirectly, I should thank Ronald Reagan, since he's the one
that started the long downward pressure on UC budgets.
I think, therefore, that for stimulating new discoveries and finding
people who have a "knack" for distributed computing, the "budget free, no
supervision" that is inherent in cheap equipment on your desk is an
essential part of the process.
>that is, it's already available.
>Beowulf mailing list, Beowulf at beowulf.org sponsored by Penguin Computing
>To change your subscription (digest mode or unsubscribe) visit
More information about the Beowulf