The two snapshots are created in response to "auto-recovery" done by Hyper-V VSS writer during VM backup.
Hyper-V writer creates a software of snapshot of all the VHDs inside the VM (using system provider). This is captured in the first snapshot on the storage system. The applications and OS in the VM is not frozen when the snapshot is created on storage system. So, later, VSS asks us to mount the first snapshot and the writer then does auto-recovery to revert the hardware snapshot to the software snapshot taken earlier. When auto-recovery is complete we take the second snapshot (_backup suffix).
Hence, the second snapshot should always be used for "connect" operations. The different between the two snapshots is essentially the auto-recovery changes.
I also wanted to check if you are running an evaluation copy of SMHV? The released version is available on NOW.
During Autorecovery VSS asks VSS hardware provider to mount the first snapshot(backing).
Snapshot mount is done with a LUN clone created in the active file system. During the mount the original LUN is renamed as lun-name_exclude and the LUN clone is renamed as LUN.
The VSS writer then does auto-recovery to revert the hardware snapshot to the software snapshot taken earlier. When auo-recovery is complete we take the second snapshot (_backup suffix). After the second snap is complete, LUN is renamed back to the original name.
So in short _exclude is the original LUN in first snapshot. The autorecovery phase renames the original lun to "Lun-name_exclude" so that it will be excluded
for the ndmp dump utility. The clone created for autorecovery will have the name of the original lun.