Rapid Recovery and VMware disk provisioning (4039683)

Depending on how the virtual disks are provisioned within VMware determines how they are presented to Rapid Recovery. This also determines how they are treated within the product as well. Unlike agent-based backups Rapid Recovery does not have a presence inside the virtual machine to gather data on the disks themselves, therefore Rapid Recovery relies on the information presented from VMware to display/gather data on the virtual machines. Below are the different types of VMware disk provisioning:

• Thin Provisioning - VMware will provision the full amount of storage specified to the disk, however it will only allocate (use) the blocks that actually have data written to them (therefore a VM only consumes space in the datastore once those blocks have data in them)
• Thick Provisioning - VMware will both provision the full amount of storage specified to the disk and allocate that much space as used blocks to the datastore (therefore the datastore free space is immediately reduced by the amount provisioned)

There are 2 additional types of Thick Provisioned disks:

• Lazy zeroed - the blocks are allocated to the VM and the space is consumed in the datastore, however data is not written to the blocks and therefore not reported as an 'active' block.

• Eager zeroed - the blocks are allocated to the VM and the space is consumed in the datastore, additionally the blocks report as 'active' (thus the blocks will appear as though there is data consumed within that block). This essentially means that the entire .vmdk reports to VMware as used/active data.

The provisioning mode affects Rapid Recovery in the following ways: (keep in mind this has no bearing on the functionality and/or use of Change Block Tracking (CBT))

1. Thin Provisioned Disks - The information provided from VMware on the disk size and the used space should be presented to the Rapid Recovery core and displayed accordingly in the Summary tab. The correct sizes and used space of each disk should be accurately displayed and base image backups should only be the size of the used space within the .vmdks as provided from VMware.

1. Thick Provisioned Lazy zeroed - The behavior is very similar to Thin Provisioning. The information provided from VMware on the disk size and the used space should be presented to the core and displayed accordingly in the Summary tab for the agent within the core console. The correct sizes and used space of each disk should be accurately displayed and base image backups should only be the size of the used space within the .vmdks as provided from VMware.

1. Thick Provisioned Eager zeroed - Eager zeroed disks present every block in the .vmdk as active, and thus Rapid Recovery will have to scan the entire .vmdk in order to determine which blocks have data stored inside them. Within the core console the disks in the summary tab will always show as full since all blocks will report as active. A base image backup will be the full size of the .vmdk. 

As stated previously this does not change the behavior of incremental backups as they are dependent upon VMware's Change Block Tracking functionality which tracks changed blocks, not active blocks. Once a base image is taken, regardless of the provisioning mode, incremental backups will be taken in the same way.