Archives


- Beowulf
- Beowulf Announce
- Scyld-users
- Beowulf on Debian

[Beowulf] Capacity / Capability Computing

Many of your questions may have already been answered in earlier discussions or in the FAQ. The search results page will indicate current discussions as well as past list serves, articles, and papers.

Search

Mark Hahn hahn at mcmaster.ca
Tue May 20 09:08:05 PDT 2008


> Definitions please.

they're lingo, and bad lingo at that, since the normal English meanings
don't give much of a clue.  some vendors like to talk "stale up vs scale
out", which is just as bad.

let's agree to use conventionally meaningful terms. 
for instance, when people say capacity, I think they usually mean 
"primarily serial or mock-parallel (embarassingly parallel, loosely coupled)"
when they say "capability", they mean "large, tightly-coupled".

come to think of it, "throughput" is another inherently poor term, widely
used.  I think people usually mean "lots of jobs per hour", but what they 
really mean is "lots of small jobs", since a "capability" cluster will
have high throughput as well, just bigger, fewer chunks...

I think it's pretty uncommon for people to use "capability" applied to 
large-memory serial machines, or fast-storage clusters.  the term has 
connotations of "grand challenge" problems, for whatever that's worth ;)



More information about the Beowulf mailing list