When planning for data recovery and verification, you should also remember that vPower NFS was not designed to deliver high throughput but to be your "spare tire" in case of failure — and to allow the machine to be brought online quickly, with minimal impact on end users.
The primary factors influencing vPower NFS throughput are as follows:
•Disk I/O latency
•Network latency between ESX host and vPower NFS proxy, as well as between vPower NFS proxy and repository
If you want to get the absolute maximum performance out of vPower NFS, you should backup to a physical server with fast local disk (15K SAS work great, but some SSD are even better), and that machine should be both the repository and the vPower NFS server. That keeps the latency down to the minimum possible level.
Overall, to get the best performance out of SureBackup jobs, FLR, and Instant Recovery, consider the following:
•Besides Windows-based backup repository servers, Veeam vPower NFS Service can run on any Windows server you choose, including the Veeam Backup server itself. However, in this case, performance may be much lower, because instead of a direct connection between the ESX(i) host and the backup repository, the connection will be split into two parts: ESX(i) host to NFS server and NFS server to backup repository.
If you store backups on a Windows-based repository, it is highly recommended to enable the vPower NFS Server on this repository (so the vPower NFS Service will run on the managing Windows server).
•If your backup proxy is a VM with the HotAdd access to the source datastore, during the full VM restore, Veeam Backup & Replication will use the Virtual Appliance transport mode to write the VM data from the backup proxy back to the datastore. This speeds up the restore process and reduces the load on the network.
•Using high-performance storages as repositories is a recommended approach. The VM boot-up speed will degrade on a slow storage, so make sure that the speed of your backup repository will support your RTO.
•The closer your storage is to the vPower NFS server, the faster your VMs will come up. Local storage is best if available.
•Make sure your vPower NFS server has sufficient RAM. For example, for booting more than 2 VMs it is recommended to have at least 6-8 GB of memory free if you plan to perform a restore (which consumes memory as well).
•Be careful if planning to recover from files stored on deduplicating appliances. They often have decent write speeds, but trying to access the full disk all at once (to start a VM) can cause severe latency. This happens due to the files that have to be undeduped and presented to the Veeam server (to be then presented to VM).
•The VMs will also consume resources on the host you are booting them up. Thus, avoid choosing a host that is already (over)utilized — otherwise, will not only the SureBackup VMs, but also the rest of the production VMs may suffer from poor performance.
•Keep in mind that the VMs are booting from the backup files themselves, so the more points your VM goes through, the slower its boot-up is.
For instance, using a Forward incremental backup that has not had a synthetic full for a month, or a Reversed incremental from 2 weeks ago, without an active full in between, will require Veeam to access more files, slowing down the process.
If your organization requires faster restores of multiple critical VMs, consider replication instead.