<

Server 2016 lets you split DHCP ranges. As I have different VLANs and ranges per building, I can give the building primary (the close one) most of the range. I do this by making the VLAN on the machine at the location have no delay, and then put in a delay in that VLAN for the other DCs. Even in a single location situation I would recommend a delay on the BDC. This allow one machine to handle normal logons, and allows you a way to gauge your network. I have a 1ms delay on the BDC and it gets about 5% of the logons. This is excellent feedback that the network is running well, and is healthy. If I had a 1ms delay and 40-60% of the logons were on the BDC, I could have an issue.
I have a 10G network with all workstations on 1G connections, including the wireless APs. The APs are AC and can handle 200 clients, with an AP in every room. I also have a single location situation with 1200 devices (plus student and staff BYOD connecting as well.)
Back to logon delay. I would highly recommend playing with this to find your network sweet spot. Find the __ms setting that results in a 10% or lower fallback to the DC that is secondary (or tertiary). If every DC is primary on a different VLAN (the primary VLAN for the physical location) then you have fallback for heavy logon times while maintaining the fastest speeds.
With network bandwidth becoming more an issue every day, it is our responsibility as IT professionals to make the user experience and fast and flawless as possible. We impact the business at hand, and possible loss of production, more than some realize. Finding the sweet spot for network logons, file access, and internet access, is one the primary ways we can make the things we do in the background obvious to those we support.]]>