How many physical servers for a Management Domain deployment with a storage appliance?

[StoopidMonkey]

Right now we're running a pair of all-in-one Veeam B&R servers plus Enterprise Manager for our environment of standalone Hyper-V hosts, also utilizing Wasabi for offsite copies. We are planning to migrate to an HPE Alletra dHCI platform running on VMware. This introduces a storage appliance and at least one StoreOnce backup appliance. We'd also like to rebuild our Veeam environment to run in a Management Domain instead of our Production Domain. Finally I'd like to utilize SureBackup and optimize things for faster Instant VM Restores, so lots to consider.

Here's what I gathered so far:

• HPE StoreOnce appliances are great for onsite archival storage but aren't good to run Instant VM Restores or SureBackup from since they don't have the IOPS for it. Typically Veem would call for a two-tier solution where backups are initially stored on a fast-disk repository and then offloaded to the StoreOnce for onsite retention

• HPE Alletra storage appliances can create short-term storage snapshots that Veeam can utilize for Instant VM Restores and SureBackup, eliminating the need for a separate high-speed repository

• Creating a Management Domain would seem to require at least two physical servers not included in the VMware dHCI node cluster. Server 1 would be a hypervisor hosting VMs of the VBR Server, Enterprise Manager, and a Domain Controller. Server 2 would be a dedicated physical Windows host for the Veeam Backup Proxy, Gateway, and Mount Server roles. Optionally perhaps a small Linux Hardened Repository server to store backups of the Management Domain environment unless I decide to route that directly to the StoreOnce appliance. Finally in the production domain I'd have to create a Veeam Guest Interaction Proxy server.

Does this sound like the right way to do this, and are the services bundled in the correct way? Our environment is relatively small with less than 100 VMs that need to be backed up.

[david.domask]

Hi Chris,

A few clarifications/comments on the above:

> HPE StoreOnce appliances are great for onsite archival storage but aren't good to run Instant VM Restores or SureBackup from since they don't have the IOPS for it.

Generally correct -- it is possible to run random IO intensive operations on a Storeonce (Instant Recovery, File Level Restore, anything that would have a random IO pattern) as long as it's understood that deduplication appliances are optimized for sequential reads. Both Veeam and the dedupe storage vendors have done a lot to overcome this limitation and you can get "suitable" instant recoveries, but it will typically never be as performant as other repository types. Usually running the random IO intensive operations from a faster block storage like you're discussing is best practice.

> HPE Alletra storage appliances can create short-term storage snapshots that Veeam can utilize for Instant VM Restores and SureBackup, eliminating the need for a separate high-speed repository

This is correct, however keep in mind that storage snapshots (arguably) are not backups and come with some restrictions. Backup from Storage Snapshot to another repository type like you are thinking and then backup copies to the Storeonce for archiving is an extremely common setup that works well, with orchestrated storage snapshots as an additional recovery option.

Overall the plan looks reasonable for the workload and I think you're covering your bases. Be sure to confirm the Gateway server specs in advance as gateways for repositories need additional resources (as linked), and this is often overlooked. Our Veeam Calculators can help with sizing estimates.