[Beowulf] Rackable / SGI

[Mark Hahn]

> involved with Linux, and open source things such as XFS we would not
> have the enterprise-level features that we see now.

unclear in several ways.  for instance, linux has hotplug cpu
and memory support, but I really think this is dubious, since 
there's damn little hardware that supports it, _anywhere_.
it's more of a "bank" feature rather than merely gold-plated 
"enterprise".

XFS may have been fairly "enterprise" for its time - it's been 
available for linux for quite a while, I think.  but if you look
at options today, is it clearly the only "enterprise" choice?
certainly not - ext3 and 4 are certainly viable, though perhaps 
not in every possible application.  JFS is presumably also an 
example of big-corp contributed "enterprise" software, but I'd 
say has had even less of an effect.  dare I mention advfs, which 
has now been open-sourced?

from my position, XFS was a semi-fringe option for people who 
distrusted ext3 for some reason.  (and there were a few solid 
ones, mainly just >8TB.)  going forward, I expect to use ext4
and probably btrfs; I don't see a lasting impact of XFS.
if IBM did buy Sun and made an effort to get ZFS Linux-ized
(Linus-ized), it would be interesting.  especially if they 
also did so with Lustre.

> And enterprise
> level features need the hardware - it is no good some geek imagining
> what would happen on a 1024 processor 64 bit system as he compiles up
> the kernel pathches on his laptop.

cheap shot - don't you remember that linux was 64b quite early because 
DEC dropped an alpha on Linus?  it was also SMP-aware pretty early.
this is not to say that all the numa features existed at the time,
though they're all pretty obvious (node-aware memory management,
care taken with per-cpu features and layout, etc).  I'm not denigrating
the big-iron contributions, but it's certainly not a "wise mature
enterprise gurus show pencil-necked linux punks how to do it right" ;)

regards, mark hahn.