[Beowulf] SGI to offer Windows on clusters
Robert G. Brown
rgb at phy.duke.edu
Sun Apr 15 12:25:10 PDT 2007
On Sun, 15 Apr 2007, Ed Hill wrote:
> Help [in the form of new packagers, new packages, package reviews, bug
> reports, etc...] is always welcome. Even as the list of Fedora packages
> grows it is still remarkably small compared to the universe of freely
> re-distributable (and thus eligible) software.
Much as I like FC 6 -- by far the best linux release I've ever used,
huge, almost totally automagical and transparent -- I've been fairly
brusquely corrected on thinking otherwise on a number of occasions, so
now I tend to be pretty cautious. FC 6 has something like 6500 packages
visible to yum on my system, but from what I understand there are
something like 20,000 accessible from Debian these days.
The impressive thing is the number of scientific and technical packages
that are making it in. I truly think that inside a year, perhaps two
(one or two more FC releases) FC will be a beowulf-in-a-box on the one
hand, and a "scientific linux" on the other, in the sense that all the
important packages required to support either one will be there. It's
already pretty close.
I'm not certain that I "like" the idea of fully integrating extras into
the main core, though. The problem with FC+extras+updates even now is
that 6500 packages is a bit difficult to get a human brain around,
especially when they are basically unsorted and in only two or three
enormous repos. Yum supports the notion of segmenting into multiple
repos, and I honestly think that it would be a good idea to take
advantage of this and have e.g. a separate "fedora games" repo, a
"fedora scientific" repo, a "fedora office" repo and so on. Flat is
good for certain machine driven automatic things -- flat is not so good
for human cognitive things.
That's why department stores are a wee bit easier to shop than junkyards
or flea markets, why libraries or new bookstores are easier to shop than
book bin used stores -- you know where to look for something. USING
packages for non-automated system configuration is all about shopping.
Yes, one can "in principle" to searches and limited sorting using yum or
toplevel tools, but I haven't found it to be as effective as all that.
At 20Kpkgs, Debian has gotten to the point where friends of mine who are
Debian users no longer bother to even try to figure out what is in it.
They install the minimum, and rely on word of mouth or blind luck to
clue them in on really nice packages of use to them in that vast list.
I sympathize. It took my two or three days to just read through the
package descriptions for "everything" in FC6 when I upgraded (and they
thought I was crazy for doing that much). With Debian it would take
over a week.
Categorical segmentation isn't a bad idea to TRY to keep things
manageable here. Whether it occurs at the repo level or elsewhere, a
hierarchy of sorts is desireable.
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