The second change affects the backup application vendors, and it has no direct end-
user impact at all. There is no concept of “differencing” backups with Hyper-V host-
level backups of VMs. Instead, a complete backup is taken. Previously, this would
result in the complete VM data set being sent over the network to the backup target
for every backup—meaning a huge amount of network traffic, huge amounts of
backup storage, and a very inefficient backup. Each vendor had to implement their
own filesystem filter that would track the blocks that changed between backups,
enabling only the delta data to be sent and stored. This was a major undertaking for
each backup vendor and could lead to performance degradation and even instability in
the event of a bug in the third-party filter. In Windows Server 2016, Microsoft has
implemented Resilient Change Tracking (RCT), which tracks the blocks changed with
VHDX files as part of the VHDX file itself, meaning that the tracking moves with the
VHDX in the case of VM and/or storage migrations. This removes the need for backup
vendors to maintain their own filter and track the changed blocks. Now via RCT, the
backup vendors can simply request which blocks have changed and back those up.
RCT requires that the files are VHDX and that the VM is version 7 or above; if not, the
old style differencing disks will be utilized for the backup.
No matter what other technologies are used, such as replication, multiple instances of
applications, or even snapshots, none of these are replacements for backups. Backups
should always be taken regularly for complete protection for all types of failure.