On the replica side, a major change has occurred between Windows Server 2012 and
Windows Server 2012 R2, related to both the amount of storage and the IOPS. In
Windows Server 2012, the IOPS requirement on the replica storage was between one
and five times that of the workload, a big range. Windows Server 2012 R2 changes this
to between one and two times that of the workload, which is a huge improvement and
is enabled by the new mechanism to manage changes to the storage. The modification
in applying the changes also reduces the storage required if recovery points are used.
In Windows Server 2012, each recovery point would be around 10 percent of the size
of the base VHD, which could be very large. In Windows Server 2012 R2 and above,
each recovery point is equal to the actual change between the recovery points, which is
a huge improvement.
Ascertaining the rate of change for each virtual machine is the key factor for all of
these estimates. To help with this, Microsoft has created a capacity planner available
at www.microsoft .com/en-us/download/details.aspx?id=39057. Once the capacity
planner is downloaded and installed, the application should be executed.
The tool asks for the primary Hyper-V server or the broker if it’s a cluster, and the
replica Hyper-V server or the broker of the replica cluster. It will also ask for the
estimated WAN bandwidth and a duration for collecting data. Next the tool will
display a list of virtual machines running on supported storage such as local and CSV
and allow you to select the specific VHDs on which you want to gather information on
the change, as shown in Figure 8.9. The tool will then monitor those VHDs for the
period of time you specify. Once collection is complete, it will know the rate of change
of the VHD and ascertain the actual processor, memory, network, and storage impact
for using Hyper-V Replica with those VHDs. A full report is generated that shows the
before and after impact to help you plan for the Hyper-V Replica implementation; I
definitely recommend using it.