New Veeam Server Build Architecture Questions

[DavidAno]

Hello Everyone,

We are building a new Veeam server and I would like your opinions / suggestions.

Background: We currently have 3 locations, all 3 locations have their own Veeam backup server onsite. For offsite backups we have cloud backups; however, they are file backups only of the source VMs, we don’t have any copies of the Veeam backups offsite. I have been given a small budget to set up a new server in a Colo facility offsite so that we will finally have an offsite copy of our Veeam backups

Site 1 - JBOD server, bare metal windows install on the server, 20TB of backup storage
Site 2 - Veeam server is a VM on our production server, backup target is a NAS , 80TB of backup storage
Site 3 - Physical Server with JBOD, Running Vmware, Veeam server is a VM, backup storage is virtual disks added to the Veeam VM, 90TB of backup storage

Disclaimer - I inherited all 3 of these solutions and have not been approved the budget to re-implement / standardize.


Consideration 1: We are a 100% Vmware operation for our production environment. In a DR scenario I would like the ability to spin up the VMs from backup and run at the offsite Colo (instant recovery)

Consideration 2: We will need 200+ TB of storage, and have budget of less than 15k.


Question 1: What would be the recommended architecture for this solution
A: JBOD Server running bare metal windows & Veeam, backup storage added direct to windows. Separate VMware server for instant recovery DR
B: VMware server, run the Veeam sever as a VM and have the backup storage attached as virtual disk(s)
C: Vmware server, run the Veeam server as a VM, but have a NAS / Separate storage appliance as the backup target

Question 2: Would any of the architectures above be better/worse for deduplication, or would it not matter when using Veeam deduplication?

Question 3: Vmware has a virtual disk (VMDK) limit of 62TB. If using architecture option C: would it be an issue to span a single partition across multiple VMDK files to achieve a single 200TB volume?

Thanks in advance!

[Mildur]

Hello David

Consideration 1: We are a 100% Vmware operation for our production environment. In a DR scenario I would like the ability to spin up the VMs from backup and run at the offsite Colo (instant recovery)

I suggest to have a look at Replica jobs. This allow you to have replicas of your VMs ready to poweron if a disaster happens. After you have solved the issue, you can failback from the backup console with all changed to the original environment. Additional licenses are not required. The same license is used for backup and replicas.

Question 1: What would be the recommended architecture for this solution
A: JBOD Server running bare metal windows & Veeam, backup storage added direct to windows. Separate VMware server for instant recovery DR
B: VMware server, run the Veeam sever as a VM and have the backup storage attached as virtual disk(s)
C: Vmware server, run the Veeam server as a VM, but have a NAS / Separate storage appliance as the backup target

If you can reconfigure all sites to be managed by a single backup server, I would go with Option A and use the server as a linux hardened repository (one hardened repository can only be used by a single backup server). This will give you immutable backups. The backup server in your Coho Site can manage components from all three sites.
The separate VMware server will be used for a VM with the backup server and as a target for the suggested VM replicas. Having the backup server in your remote side allows you to manage replicas in case of a disaster without restoring configuration backups first.

All other options do not protect your backups against accidental or forced deletion of malicious actors. This is something I recommend all customers this days. Every investment in backup strategy must consider air-gapped or immutable backups.

Question 2: Would any of the architectures above be better/worse for deduplication, or would it not matter when using Veeam deduplication?

If you use a hardened repository with XFS file system, you will be able to use FastClone aware backups.
FastClone allows you to run spaceless and fast synthetic fulls. Instead of writing a copy of all unchanged data blocks for new full backup, the backup job will reference to unchanged blocks of previous backup files when creating the full backup file. A lot of space saved.
https://helpcenter.veeam.com/docs/backu ... ml?ver=120

Question 3: Vmware has a virtual disk (VMDK) limit of 62TB. If using architecture option C: would it be an issue to span a single partition across multiple VMDK files to achieve a single 200TB volume?

We don't recommend virtual disks as backup repositories. If your ESXI Host has an issue, you won't be able to read backup data until you have rebuild the ESXI host. Second, an malicious attacker with administrative access to the vCenter can just delete the backup VM and your entire backup data on that virtual disk is gone.
But to answer your question, Veeam will be able to work with such volume. Or you use single 62TB volumes as performance tier extends in a SOBR.

Best,
Fabian