When a virtual machine crashes, there are two ways to quickly recover it – first is by using the VMware snapshot copy and second is by restoring an image-level backup.  Most VMware environment though do not usually perform snapshots on the virtual machines (VMs) due to increased usage on the primary storage, which can be costly.   On the other hand, using traditional method to restore image-level backup can take longer since it has to be copied back from the protection storage to the primary storage. 

However, most backup solutions nowadays – including Netbackup, Avamar/Data Domain, and Veeam – support VMware instant recovery where you can immediately restore VMs by running them directly from backup files.  The way it works is that the virtual machine image backup is staged to a temporary NFS share on the protection storage system (e.g. Data Domain).   You can then use the vSphere Client to power on the virtual machine (which is NFS mounted on the ESXi host), then initiate a vMotion of the virtual machine to the primary datastore within the vCenter. 

Since there is no need to extract the virtual machine from the backup file and copy it to production storage, you can perform restore from any restore point in a matter of minutes. VMware instant recovery helps improve recovery time objectives (RTO), and minimizes disruption and downtime of critical workloads.

There are also other uses for instant recovery. You can use it to verify the backup image, verify an application, test a patch on a restored virtual machine before you apply the patch to production systems, and perform granular restore of individual files and folders.

Unlike the primary storage, protection storage such as Data Domains are usually slow.  However, the new releases of Data Domains have improved random I/O (due to additional flash SSD), higher IOPS and better latency, enabling faster instant access and restore of VMs.