Linux vs FreeBSD clusters (was: how are the Redhat product changes affecting existing and future plans?)
Andrew M.A. Cater
amacater at galactic.demon.co.uk
Fri Nov 7 16:36:48 EST 2003
On Fri, Nov 07, 2003 at 02:17:48PM -0500, Mark Hahn wrote:
> why? I've never understood the evangelical aspect to *BSD
> (or for that matter Debian). is there a tangible, measurable benefit?
For some of the list this is old news :( Sorry, but as the co-author
of the Distributions HOWTO (and I know it badly needs updating :) ),
I can't let this pass.
In the beginning was chaos: build your own Linux machine from random
bits and pieces, bootstrapping from Minix if necessary.
Soft Landing Systems (SLS) introduced the whole concept of a Linux
distribution - a collection of code more or less known to work together.
In 1993, Patrick Volkerding got fed up with the problems of SLS
and founded Slackware. Coincidentally, within a couple of months, a
university student named Ian Murdock did almost the same and founded
Debian, named for himself and his then girlfriend Debra.
Mark Ewing and the Red Hat founders came along a little later, liked
some of the concepts from Debian and the idea of package management
and introduced the Red Hat Package Manager which spawned .rpm packages.
SuSE came out of Slackware via Unifix and Jurix and adopted .rpm's a
little later. Mandrake was a French attempt to localise Red Hat and
introduce some better packages ... and so on.
Don Becker can probably back me up on this - the first cluster was
a quick project to fill a specific need for NASA. Cheap, commodity
hardware and a quick win. It's name was Beowulf (for the mythical
hero). It wasn't intended to go further than NASA and be a short term
thing. But it snowballed to the point where everyone wanted "a Beowulf"
The first Extreme Linux disk was based on Red Hat 5.x because that was
what happened to be around at the time. I've still got mine somewhere -
a quick bootstrap to experiment with a cluster (in the days when 4 x
486's still counted for something).
Much the same with Extreme Linux 2 and the (semi-commercial/commercial)
Scyld. Then commerce woke up to and "commoditised" clusters. Clusters
don't have to be Red Hat - nothing Linux _has_ to be anything - but
many of them are.
There have always been various distributions: various clustering
solutions/hardware have come and gone. Everyone has "their" cluster and
"their" problem/solution set. It may still be quicker to build your
own minimalist system / cheaper to use cast-offs / more economic for
you to use ultra-high performance interconnects and networking - that's
for you / your budget holder / sysadmins / vendor / cooling plant
vendors to decide :)
I advocate Debian not just because I use it a lot (for the record, I've
also used Red Hat 4.2/5.*/6.*/7.*/8/9/Fedora beta, Mandrake, SuSE,
Slackware and the late lamented Linux-FT) but because it has some good
qualities, runs on lots of hardware platforms and is relatively
unencumbered by nasty legal agreements/high fees. That's my choice -
I'll happily help others to run it/port programs so they fit in the
Debian distribution/ask vendors nicely if they'll consider support
for Linux. It may not be your choice or the choice of others -
but it is always worth trying stuff out and being open to change.
[Rabid flame mode on: For the record, Debian, despite being maintained
by an army of multi-national, multi-lingual volunteers who occasionally
manage the semblance of close formation, is _not_ a niche OS - some
figures put it second to RH in terms of popularity as a Linux
distribution :) Go and write GNU/Linux 1000 times as a penance.]
As far as *BSD goes: The BSD's have a longer pedigree.
Some people swear by (others swear at) their networking
capability. If you've grown up with SunOS and many of the
other commercial Unices, much may feel familiar. NetBSD
will run on (almost) anything, FreeBSD is less hardware agnostic
and differently focused ... and so it goes.
A lot of arguments have been thrashed out over the years which
generate more heat than light (vi vs. emacs vs. any other editor /
.rpm vs. .deb / apt vs. yum vs. urpmi vs. up2date :) ) and this
is probably one of them, so it's probably not worth a massive OT
thread to follow up.
> I'm not sure it's effective to advocate a niche OS/dist for
> ideological reasons or just plain personal preference...
See above: your niche OS may be someone else's ideal. An awful
lot of niche Linux variants have tried to set themselves up as
"the standard" - and vanished without trace.
Beowulf mailing list, Beowulf at beowulf.org
To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
More information about the Beowulf