And uptime is important, so you want to have that secondary running and ready, with a proxy in front of everything so you can switch as soon as you detect a failure. That's three hosts, plus your alerting has to be separate too, so that's four. Now, to orchestrate all this, we'll first get out Puppet...