[AUUG-Talk]: Proprietary Unixes (Dead?)
Gary R. Schmidt
grschmidt at acm.org
Thu Oct 4 12:21:07 EST 2007
Enno Davids wrote:
> datacentre floor any time soon. No. Can you configure me a box with >100
> cores on an open source platform today?
Yes. SGI Altix boxen. You want 10,240 cores? No problem. Runs SLES
(and RHEL, IIRC).
> Can you split it in two or three
> for now and re-arrange its resource allocation in the event of a disaster?
Okay, do you want an ICE system? New idea, lots of blade-ey type
things, running domain-ey type problems. (I have *no* idea what I mean
here (I still haven't managed to connect to one of them, I was supposed
to do some sort of QA on it), go read the blurbs on it, just don't
fixate on the bit where the blades have no local disk.) Runs SLES.
(May run others, I slept through the presentation.)
> So proprietary UNIX dead? Not even close.
Damn right. The Linux stuff is fine for the beyond-the-bleeding-edge
stuff, where the prospect of down-time is offset by speed and cost (APAC
supercomputer, frex), and for tightly controlled stuff (how many NAS
heads are running a Linux variant, or mobile 'phone, or embedded thingie...)
When it comes to *reliability*, long-term, 24x7, five-9's and all that
crud, you go to the high-end *hardware*, and then, after you've spent
the money, say, "I'll run so-and-so's Linux distro," and the hardware
supplier says, "That's nice, but we can't support you at five-9's for
that, you're on your own," you'll quickly swing back to the proprietary
OS. And if you don't, the board will shortly have your head.
P.S. And yes, IRIX is dead, the last version was 6.5.30, there's a
collection of (mutually antagonistic) patches that are referred to as
6.5.31, but that's just to increase confusion.
Armful of chairs: Something some people would not know
whether you were up them with or not
- Barry Humphries
More information about the Talk