From: Eugene Leitl (eugene.leitl@lrz.uni-muenchen.de)
Date: Fri Jul 28 2000 - 19:24:32 PDT
Damien Morton writes:
> Forgive me, Im a little drunk, and not at all religious when it comes to
> OSs. If my T1 connection has 99% uptime, what do I care if my cluster has
> 99.9% or 99.99% uptime.
Clusters running a single application have to roll back to last known
good state. They're usually only connected to the local network. You
cannot ssh login to most installations from home -- quite a few
systems which allowed that got hacked.
> What is the cost difference (in hardware, software and personel) between a
> solution that offers 99.9% uptime and 99.99% uptime. Does that cost make
> sense in terms of my business?
You should read the Beowulf archives on why NT is such a loser for
scientific computing on clusters. If we're talking about message
latencies in 35 us range, do you think NT can utilize that?
This archive was generated by hypermail 2b29 : Fri Jul 28 2000 - 20:29:35 PDT