We do it based on 24 hours for critical items on a complete failure (this gives us time to bring another server up and restore to it), 48 hours for required funtionality that is not critical and 1 week of bits and pieces crap (we call it non-essential functionality ). Deleted data restores are 24 hours and we charge a fee in some cases. Overall most of the systems I cover have no major impact on the corporation as a whole.

But I have seen really poor SLAs of say 99.99% uptime guaranteed on systems that have no failover point so when they go down for 5 minutes we give them as much grief as possible since they forced us to rely on some of these systems for authenication (poor implementation of a good design).

The key is how important is it and how quickly can you actually get it back up. If critical make sure you find a method to keeping it up or getting back up that shows it's importance, but make sure your implementation supposrts what you promise.

Edited by - antares686 on 12/03/2003 04:24:15 AM