> based on your figures, a system that's 99.99999% accurate would still make
> more than a quarter million mistakes each day (1 per 100,000 transactions).
> increasing the reliability exponentially increases the costs exponentially
> as well.
Agreed in part: but they have to solve this problem anyway, with or
without solving their Y2K problems. I'm not sure I accept that
billions of test cases are necessary, however.
> > I wonder what it would cost to rip out their entire computing
> > infrastructure and start over.
>
> Probably 10-20 times that much. Hardware, software, conversion,
> cut-over, lost revenue; the smallest of these costs is hardware.
Hardware is now dirt-cheap. Software? Cheap and getting cheaper,
especially if you use as much freeware as possible. Conversion? Set
up the new stuff in parallel -- including an entirely new network -- test
the hell out of it, make the new UI asymptotically close to the old one
to minimize training costs, and throw the switch. (Yes, I have actually
done this, on a live network with 60K users. It worked. No big deal.)
And yes, there are hardware solutions for this sort of thing,
like the Sequent NUMA machines, which are fault-tolerant, scalable,
and deliver obscene amounts of processing power. I wonder if anyone
in the Linux community has tried clustering these? 1/2 ;-)
>> ... I'd love to take on this job
>> with a proviso in the contract that I get half the savings off
>> their current projected budget. ;-)
>
>You and what army?
Nope. No army. Trying to attack this problem with an army would
be mistake #1. A relatively small group would have a much better
shot at it.
---Rsk
Rich Kulawiec
[EMAIL PROTECTED]
____________________________________________________________________
--------------------------------------------------------------------
Join The Web Consultants Association : Register on our web site Now
Web Consultants Web Site : http://just4u.com/webconsultants
If you lose the instructions All subscription/unsubscribing can be done
directly from our website for all our lists.
---------------------------------------------------------------------
