Since DMA failover setups are for hardware redundancy, is a failover setup still beneficial in case the DMAs are all running on VMs?
And going a bit further, if I'm planning to migrate my current DMS to a virtualized environment about a year from now, should I go for the 'traditional' failover setup or should I count on the new swarming?
I think it is currently still beneficial, as in a failover setup both failover agents are continuously synchronizing all files, settings, and the databases.
While virtual machines allow you to spin up a replacement machine in seconds, the challenge is in the synchronization.
Starting a VM from a previously created image may not contain the latest state of the DMA, because it is not continuously synchronized.
Using the same VHD file as the original VM still exposes you to hardware failure of the physical disk(s) containing the VHD.
In the future swarming will indeed provide an alternative, as then the settings are synced within the cluster and the data is in a distributed Cassandra and/or Elastic cluster. This feature is currently still in development.