Re: Enquiry regarding Linux in Mission Critical situation

From: Michael Heiming (
Date: 06/29/02

From: Michael Heiming <>
Date: Sat, 29 Jun 2002 01:22:05 +0200

Christopher Browne (<afi2aa$enhuk$>):


> That means _never_ having an opportunity to change kernels, even
> if that would seem worthwhile.
> Every extra 9 that gets added is likely to exponentially increase
> the cost of the system, as you will absolutely _need_ massive
> amounts of redundant hardware.
> For instance, if you can keep a pair/trio of systems synchronized
> against one another, it may be possible to take one down, and have
> the others undetectably take over its duties in a tiny fraction of
> a
> second. That would allow nodes to be individually upgraded or
> repaired.

The sync is kind of a problem for many apps, there are some
features like clusterwide shared fs missing, even if there are some
commercial solutions available.

Another great thing would be process migration from one cluster
member to another. AFAIK there is no *nix which can do this today.

Tru64/TruCluster is rumored to get this feature, which is IMHO the
most advanced *nix cluster.

Let's wait what the next generation of IA64 will offer, Linux won't
get in real mission critical computing, if you run on Wintel PC 32
bit hardware.
> But ensuring that the synchronization is done in a way that is
> transparent outside the system may not be simple or easy.
> Someone promising 10 nines has got to be smokin' something.

Or he might run a $$$$$$$$$$$$ non-stop Himalaya VMS cluster.
I've heard about VMS boxes running +10 years without reboot...;-)

Michael Heiming

Remove the +SIGNS case mail bounces.