From:	Terry Richards <trs AT idt DOT net>
Newsgroups:	comp.lang.asm.x86,comp.lang.c,comp.os.msdos.djgpp,comp.os.msdos.programmer,comp.os.os2.programmer.misc
Subject:	Re: FreeWin95 Project
Date:	Wed, 19 Feb 1997 11:16:55 -0500
Organization:	Terry Richards Software
Lines:	101
Message-ID:	<330B2777.4936@idt.net>
References:	<5ee52d$q80 AT News DOT Dal DOT Ca>
Reply-To:	trs AT idt DOT net
NNTP-Posting-Host:	206.20.33.10
Mime-Version:	1.0
To:	djgpp AT delorie DOT com
DJ-Gateway:	from newsgroup comp.os.msdos.djgpp

Norman L. DeForest wrote:
 > 
 > [ references trimmed to prevent newsreader/poster indigestion. ]
 > 
 > Terry Richards (trs AT idt DOT net) wrote:
 > : Todd A. Fiedler wrote:
 > : :> This idiot must not have to actually manage NT servers. We run a
few
 > : :> applications on our NT servers that cost us a lot of money when
not
 > : :> available. You might call them mission critical. If I installed
the NT
 > : :> 4.0 service pack 2 on my servers (assuming I was dumb enough to
upgrade
 > : :> to 4.0) and then my servers crapped out to that wonderful "blue
screen
 > : :> of death", as MANY people experienced, I would probably NOT be a
very
 > : :> happy person. It is conceivable that I would be an UNEMPLOYED,
not very
 > : :> happy person. So before you Microsoft slapheads start spouting
about a
 > : :> dozen or so bugs NOT being a big deal, maybe you should have to
stake
 > : :> your livelihood on that code.
 > 
 > : Anybody who runs apps *that* critical and doesn't try *any* upgrade
on a
 > : test machine first is on very shaky ground calling somebody else an
 > : idiot.
 > 
 > : Terry Richards
 > : Terry Richards Software
 > 
 > That wouldn't help much.  On this other machine, how do you simulate
the
 > network load that will be on the operational system.  The only SURE
test
 > would be to duplicate your entire system and install the upgrade on
that.
 > Then simulate the same load that is on the working system.
 > 
 > Of course, to do that you would need twice as many computers as you
now
 > have and would have to double your site licence to allow installing
ALL
 > of your applications on the duplicate system.  Then you need staff to
use
 > those systems to give them the same load as the working system.  Oh,
and
 > you need a few extra printers to provide for simulated network
printing
 > and you need . . . .
 > 
 > Can you spell $$$$$$$$$$$$$ ?
 > 
 >                 Norman De Forest
 >                 af380 AT chebucto DOT ns DOT ca
 >                 http://www.chebucto.ns.ca/~af380/Profile.html
 >                 (A Speech Friendly Site)
 > 
 >
.........................................................................
 > Q.  Which is the greater problem in the world today, ignorance or
apathy?
 > A.  I don't know and I couldn't care less.
 >
.........................................................................

Norman,

There are many ways to answer this. Eventually, they all come down to
"just how critical is the application?". 

You could indeed duplicate the entire system, I know of at least two
sites that do/did this with million dollar IBM mainframes.

Or you could design a fault tolerant system that normally runs on N
servers but can still provide acceptable performance on N-1 servers.
Then you upgrade the servers one at a time. I would isolate the machine
being upgraded from the rest of the network until I was convinced it was
stable. There's still a small window of risk when you connect back to
the network but it shouldn't take more than a few seconds to yank it
back out if it causes a problem.

Or, If your application is a little less critical, you could schedule
the update for off-peak or down-time.

Or, ...

The point is that there are ways of minimizing risk in any situation.
These ways may cost more than the risk is worth. Or they may not. It all
depends on how "mission-critical" the app really is. I suspect that, in
this case, the original poster was engaging in anti-MS hyperbole and
that his application isn't really that critical. If it really *is* that
critical and he is not using some kind of plan that allows for failures
then he really is walking on thin ice and, maybe, his job really should
be at risk. What if one of those servers blows a power supply?

Terry Richards
Terry Richards Software

- Raw text -