[Beowulf] Is Beowulf a standard?

Robert G. Brown rgb at phy.duke.edu
Thu Mar 15 04:32:43 PDT 2007


On Tue, 13 Mar 2007, Alice Poma wrote:

> Hi there,
> As as a newbie I have a question, "Is Beowulf a clustering standard?" if yes
> "What makes Beowulf a standard"
> As I read about Beowulf, it appeared to me as a method for starting Linux
> clustering, but some people call it a standard, I couldn't understand that
> what makes Beowulf a standard, so your help is appreciated

Beowulf is not a standard.  Standards are usually produced by some sort
of consortium of governing body and are good for certification or to
define a common working environment.

"A beowulf" is particular instance of the genera of cluster computing,
that includes grid-style, cluster-of-workstation style, and many other
kinds/styles of cluster computers.

At it's broadest "a cluster computer" is more than one CPU
interconnected and communicating with a network.  Certain multiprocessor
"single" computer designs are arguably clusters -- IBM's SP series,
AMD's current hypertransport SMP systems.  There is a lovely but dated
book by Pfister entitled something like "In Search of Clusters" that
shows how cluster paradigms are really very old in computing and
considerably predate "the beowulf".

The beowulf was "special" because it built a cluster out of commodity
off the shelf (COTS) hardware on top of an open source operating system
(linux), making it the cheapest possible do-it-yourself supercomputer,
although there were quite a few people who were doing do-it-yourself
supercomputers on top of proprietary workstations and operating systems
at the time (which were obviously more expensive but were more powerful
as well).  PVM was indeed developed starting in 89 or 90 (IIRC) to
facilitate this development and was a key aspect of early beowulfs.

Afterwards, nearly everybody in the world who needed supercomputing had
a choice -- big iron (minimum buy-in at $100K, more or less, and that
for just a very few processors in e.g. an SGI refrigerator that was,
incidentally, a cluster) or buy a pile of PC's for $1-3K each, put linux
on them, network them, and "poof" -- instant supercomputer!

   HTH,

     rgb

>
> thank you all
>

-- 
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