Comments on uptime targets - the other point I would make here is that for nearly any system, there are multiple pieces that can independently fail without taking the whole thing down, so you actually need to consider a whole bunch of separate uptime targets.