Robert G. Brown
rgb at phy.duke.edu
Sun Mar 6 06:39:06 PST 2005
On Thu, 3 Mar 2005, Trent Jarvi wrote:
> Just a heads up.
> This page appears to be corrupted.
> While not all Beowulf clusters are supercomputers, one can build a Beowulf
> that is powerful enough to attract the interest of supercomputer users.
> Beyond the seasoned parallel programmer, Beowulf clusters have been built
> and used by programmers with little or no parallel programming experience.
> Beowulf clusters provide universities, often with limited resources, an
> excellent platform to teach parallel programming cNvq0ZhTgBrP
Corrupted and out of date, too:-)
Nobody who looks at the top500 list (whatever my opinions about its
basis;-) would nowadays say that one can "build a Beowulf that is
powerful enough to attract the interest of supercomputer users".
It's getting to be much more of a "seasoned parallel programmers (a.k.a.
`old guys') can remember a time when parallel programming was carried
out on `supercomputers', basically a name for a cluster with proprietary
internal processor interconnects".
Linux hasn't finished taking over the world, although it continues to
make excellent progress with all sorts of economic and historical forces
driving it. "Beowulfs" in the generic sense of COTS clusters with
network interconnects for IPCs, pretty much have taken over the
supercomputing world with only a few exceptions, and even those
exceptions are relying less and less on anything like a custom
communications bus. Not even the engineering of the dedicated systems
scales, while using a "COTS" communication platform such as Myri or
Dolphinics, or IB or even gigE lets you leverage all sorts of useful
work done by other humans devoted to this one purpose or this purpose
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