When you configure a virtual machine for replication, vSphere Replication starts an initial configuration task during which a replica virtual machine is created on the target site, and data synchronization occurs between the source and the target vCenter Server site.

The speed of data synchronization depends on the availability of information about block allocation of the VMDK files. vSphere Replication uses this information to find empty regions of the disks and accelerate the sync operations by skipping these regions. The speed of data synchronization also depends on the site for which block allocation information is available.

If the allocation information is available at both sites, data synchronization occurs at the highest possible speed.

If the allocation information is available only at the source or the target site, vSphere Replication skips the empty regions on the VMDK disks at that site, but processes the entire disk at the site where allocation information is not available. Therefore, data synchronization is slower.

If the allocation information is not available at either site, data synchronization is done by comparing all blocks between the source site and the target site, even if many of the blocks have not been allocated on the disk by the guest OS. This is the slowest method for data synchronization.

Note

The availability of block allocation information has little effect on the speed of data synchronization for VMDK disks that are almost full.

The availability of allocation information and the degree to which vSphere Replication can use it to accelerate data synchronization depend on the ESXi versions, the vSphere Replication Management server versions, the type of VMDK disks, and the type of volumes on which the disks reside.

Product Versions at the Source and the Target Site

The acceleration of initial synchronization is supported only on ESXi hosts 6.0.x or later.

If the ESXi and the vSphere Replication Server on the source site are 6.x or later, but the vSphere Replication Server or the hosts at the target site are not 6.x or later, the allocation information will be available only on the source site.

If the vSphere Replication Management servers at the source and at the target site are both 6.x, but one or more ESXi hosts at the target site are not 6.0 or later, if the vSphere Replication Management server selects a target host that is not 6.0 or later, there will be no allocation information available on the target site.

Note

Because vSphere Replication Management server 6.x cannot select only ESXi 6.0 hosts for the initial synchronization, the acceleration of the operations might vary depending on the selected host. To achieve maximum acceleration, all ESXi hosts at the target site that act as storage servers for vSphere Replication should be ESXi 6.0 or later.

The Type of the Datastore

Disks on VMFS or VSAN datastores provide full allocation information.

NFS datastores cannot provide allocation information for the disks that are located on them.

Note

Replication disks on the source and the target site can be on different datastore types. The acceleration of the initial synchronization depends on whether both sites can provide allocation information, or only one site. If none of the sites can provide allocation information, no acceleration occurs.

The Type of Virtual Disk

Lazy zeroed thick disks, thin disks, and vSAN sparse disks, Space-Efficient sparse disks, and VMDK sparse snapshots provide allocation information.

Eager zeroed thick disks do not provide allocation information.

Virtual disks that are based on VVOLs are native to the volume. vSphere Replication 6.x can get allocation information from them only when they are on the target site. For this reason, the acceleration of the initial synchronization will be partial.