Re: -= PROPOSAL =- Release sarge with amd64
email@example.com (Andrew M.A. Cater) writes:
> This long post (in a long thread) sums up a couple of problems in
> support and stuff for amd64. I'm being asked NOW for advice and help
> by a couple of colleagues who are planning for a potential AMD64 cluster.
> They haven't bought the hardware yet but are being very sensible and
> specifying the libraries and applications they want from the cluster
> vendor (mostly math / physics libraries): they are not explicitly
> specifying a distribution. They keep asking me - "Is <foo> library / binary
> available in Linux" and seem surprised when I say "Not in Red Hat Enterprise
> Linux (or SuSE as far as I know) but it is in Debian" I run a tiny cluster
> for testing on 32 bit: they know I'm a Debian zealot but they thought
> all Linux was the same. It isn't :)
One of the Debian amd64 porters (Mattias Wadenstein, maswan) is
maintaining just such a cluster of 190 amd64 systems. I think they
(his workplace) is also doing math / physics simulations on them. You
might have similar use and should talk to him.
> I need to convince enterprise bean counters that 64 bit Linux is a valid
> idea. They'll probably want to buy Red Hat in any event- but probably don't
> know the extent of the support they'll get outside the base operating system.
> [I can't imagine RH supporting GMP library problems with enthusiasm, for
> example]. Pure64 seems to me to be an ideal solution in the short ->
> medium term: being 64 bit only will probably save the mathematicians the
> trouble of one more variable quantity :)
A enormous selling argument for amd64 (instead of intel) is that you
can run all the existing and supported i386 stuff now. But you can
also try out the 64bit stuff and change over to it if it works or when
support for it becomes available.
Support in RH and SuSe will be swiching over to 64bit completly as the
majority of customers become 64bit. With intel following in amds
footstep 32bit will soon be a minority.