In our network, we manage most IP addresses with DHCP reservations, servers, printers, video cameras, security panels etc.. about the only thing that doesn’t have DHCP reservations are workstations/laptops/devices etc..
I’ve always considered our single DHCP server a weakness as a single point of failure, and it has been (especially if you delete all of your reservations! Hey, it happened). Until Windows Server 2012, the best you could do was to split your DHCP scopes between two or more DHCP servers. The weakness of this approach is that your scopes need to be large enough to support a failure scenario, otherwise you can only support half of your devices (depending on the length and knowledge of your failure) and you still are susceptible of losing your DHCP reservations.
Worry no more. Windows Server 2012 supports true DHCP fail-over. We haven’t actually implemented it yet but will soon. We are in the process of moving to a full virtualized infrastructure with a DC being on of my only physical machines. For reasons that only make sense in my brain, I will probably move our primary DHCP to a physical server as well (or co-locate on the DC) and have the fail-over DHCP running in our Hyper-V cluster.
Here is a link from Microsoft explaining how to configure this setup:
DHCP Fail-over in Windows Server 2012