On Wed, 18 Jun 2003, der.hans wrote:

> > > I mean no offense by this at all, but be realistic:  What if your mail
> > > server smokes a power supply while you are at work/in the dentist's
> > > chair/whatever...  You're screwed.  If the Cox SMTP server tanks, they have
> > > a dedicated group to getting it back up, and I'd bet a donut (maybe two)
> > > that they have a failover system that works reasonably well.

> > I would imagine that cox could just change dns to point
> > smtp.west.cox.net to a different server to handle the load - perhaps a
> > round robin dns type setup.
>
> Well, it would make sense for them to have failover stuff in place.
> Unfortunately that hasn't been my experience with ISPs.
>
> ASU used to go down all the time. Granted, ASU has quite the complicated
> setup.

I work for a big impressive Fortune 500 company with
half a floor full of admins and wagonloads of direct
hands-on support from Microsoft.  The Exchange boxes
lose functionality (network, os, daemon, or other)
for at least an hour or two a couple times a month.

I run my own SMTP on a single plain-old desktop
machine with a 600 Va UPS (albeit colo'd).  My email
has been down for about 10 hours in the past three
years.

Big companies have big networks, big networks have
bigger problems.  But they also have more support.

But that's not the question.  The question is what
level of support is a company willing to pay for?
If they were providing 99.999% uptime, the CFO would
be (rightly) pissed off.  The customers aren't paying
for that level of quality.  If they are hitting that
level of quality (at least according to bottom line
oriented modern corporate America), they are spending
too much on maintenance.

My guess is that Cox has had a lot of experience
with finding the balance point between screaming
customers and screaming financial officers, and that
it falls well on the low side of 99.999%.