Thursday, 9 January 2014

Windows Server 2012 R2 and Two Smaller Servers Over One Big One

Having some thoughts on designing client's IT solutions to provide a relatively simple setup that allows for business to continue on in the event of a hardware failure.

Windows Server 2012 R2 gives us a few more options to facilitate business continuity.

Two smaller servers running their workloads allows for a number of different scenarios for recoverability:

  • Hyper-V Replica
    • For obvious reasons
  • DHCP Failover (built-in, run the wizard after installing the DHCP Role on two systems)
    • Very easy to do and gives clients full DHCP if one box goes down (no need to flip a switch somewhere else to enable DHCP)
    • Shares all Scope Options and Reservations between the two

Some of the benefits of this setup are:

  • AD is covered in the event of a full-stop
    • Hiccups can be taken care of by Burflags and/or AD Recycle Bin
    • AD continues despite one server going full-stop
  • File services and LoBs come back online when replica failover kicks in
  • A good backup regimen with restore tests allow flexibility (ShadowProtect)

Our preference has grown into having two key resources duplicated:
•    AD/DNS/DHCP across two separate VMs (2x servers)
•    Hyper-V Replica for VM hosting files and key LoBs

That folks is a poor man’s/woman’s "Cluster" setup.

Yes, there is a bit of extra cost involved for the licensing side of things. And, there may be a price difference on the hardware side of things.

But, when we look at the lifetime of the solution and take that extra cost we can then draw up a dollar amount per user per month using a 36 or 48 month amortization table (or even 60 month if five year warranty) and justify it as the cost of insurance relative to business stoppage costs. This works for us pretty much every time! :)

Philip Elder
Microsoft Cluster MVP
Author: SBS 2008 Blueprint Book

Chef de partie in the SMBKitchen
Find out more at
Third Tier: Enterprise Solutions for Small Business


Ben Krause said...

I see you put a lot of emphasis on redundancy with regards to servers by doing clusters etc but I don't recall seeing anything with regards to networking redundancies. What if any do you do as far as routers and switches go?

I'm not being critical, I'm just curious. :)

Philip Elder Cluster MVP said...


When one asks constructive questions then no worries. :)

For clusters a pair of switches split across a pair of quad-port Intel Server Adapters (2 ports each NIC in each switch).

Then on into the client's switches. There is ultimately noting we can do about the client side if their switch goes down. But, at least the cluster keeps humming along.

Thanks for the question,