I don't get the downtime comparison. In government, it's not the result of an unplanned outage that is being addressed to bring the system back up. Rather our governmental "sysadmins" decided to bring the system down, and can simply decide when to bring it back up.
Downtime is downtime - if you're measuring service reliability the planned or unplanned nature of it being offline doesn't really come into play. When people say five 9s, they don't mean "except for when you decide to take it offline".
What? No, that's completely untrue. We guarantee 4 nines on some of our services, but they're measured with 2 hours of allowed downtime early Sunday mornings (we usually don't need this, but very occasionally we do).
How do you figure? 99.99% of the time our clients expect us to be up, we're up. We're not the electric company; if we're up the right 166 hours in a week our customers aren't impacted.
If you turn your monitor off at night, does it suddenly have 70% availability/reliability?
99.99% availability is almost always predicated on 24/7 uptime numbers, not 24/6 plus an extra 22 hours.
If you say to someone "We have four nines of uptime", and they ask how you apply kernel patches, and you say "oh we just do it during our 1% of downtime", they'll boggle at you, and for good reason.
You're wrong. It's really up to the SLA and business circumstance. For instance, a lot of financial businesses don't operate on the weekend; surely their maintenance on Sunday doesn't count toward them operating at full uptime while open.
