Veeam Server (1 Gb uplinks) I have some questions

[rockstershine]

Hello guys,

I want to deploy Veeam Backup and Replication to backup some VMs I will host on a Nutanix AHV cluster, the backup repository will be a Catalyst store (HPE StoreOnce). I have a question about uplink speed and the impact on Backup Jobs/Restore Jobs?

So Veeam 11a will be deployed on a 1U rack server (ProLiant G7) that has 3 dual-port Ethernet adapters, that's 6 1Gbps ports. (And suppose I do some sort of NIC Teaming so I end up with 6 Gbps throughput)
On the repository side there's an HPE StoreOnce with 4 10Gb uplinks. The connection will go through Ethernet (Veeam Server connected to two switches, on which Nutanix and StoreOnce will be connected as well).

If I understand correctly when a Backup Job is launched Veeam leverages the Proxy (which I will host on the AHV cluster) to take snapshots of VMs, so this request goes through 6Gbps, then Nutanix snapshots the VMs and sends them back to Veeam on 10Gb then Veeam writes these to StoreOnce on 6Gbps once again (correct me if i'm wrong)

While the restore job will flow first at 10gbps from Storeonce to the switch then 6gbps from Veeam to Nutanix?

I sound confused about this I'm aware, I would appreciate any help with figuring this out and maybe suggestions?

Thank you

[HannesK]

Hello,
and welcome to the forums.

Independent from Veeam: 6x1Gbit is rarely 6Gbit. To get 6Gbit from six individual links, there needs to be a chance for a load balancing. Load balancing is defined by a load balancing algorithm (configured between switch and server / other switch). For example with MAC-based load balancing, there need to be enough MAC addresses for the communication.

The backup traffic flow is: Nutanix proxy running on Nutanix -> gateway server defined for the StoreOnce repository -> StoreOnce (or the other way around for full VM restore).

StoreOnce (same like other inline deduplication appliances) are known for slow restore speed (if that's the background of the question). That's why the recommendation is to use plain disk storage instead of inline deduplication appliances.

Best regards,
Hannes

[rockstershine]

Hello HannesK,

Thank you! Glad to join the community.
Thank you for the reply and clarification.
So if I understood correctly, the speed of the VBR physical server is not relevant, as the actual data movement goes from Proxy to GW server to StoreOnce and that's all a 10G flow (reversly for backup).

The background of the question is to know how does the speed of the physical adapters on the physical server hosting Windows/Veeam software impact anything?

Best regards,
Ayoub

[HannesK]

Hello,
the question is, where the gateway server is... if it is the backup server, then it's relevant (the user guide page explains how it works if "automatic selection" was chosen).

If your gateway runs for example as VM on 10G (or more) on Nutanix, then the backup server network speed is irrelevant for full VM restore (I'm only talking about full VM restore here).

File level restore goes via network (admin$ / SSH + Veeam transport). So the proxy is irrelevant.

Best regards,
Hannes

[rockstershine]

Hello,
The gateway server is going to be the same as the Proxy server
Your answer explained it all. Thank you again.