Virtual machine disk files (VMDKs) can be deployed in three different formats—thin, thick, or eagerzeroedthick. Both thin and thick disk files use lazy zeroing, where the initial zeroing of the disk blocks is delayed until the first write. With eagerzeroedthick disks, the disk blocks are preallocated with zeros at the time of disk provisioning, and it is unnecessary to zero the disk on a first write basis during normal running operations. This provides approximately a 10–20 percent performance improvement over the other two disk formats.

Because Microsoft SQL Server high availability features are highly sensitive to system response time, the additional small overhead of disk zeroing during normal operations can cause unnecessary disk latency and potentially cause a false cluster failover event. If you are deploying AlwaysOn Availability Groups on any vSphere platform, define an operational standard to employ eagerzeroedthick disks for SQL Server data, transaction log, and tempdb files to mitigate the risks associated with non-eagerzeroedthick disks.