Hi,

The calculation is a strong recommendation for calculating what you need.

Targeting an 8 hour full backup is ambitious though with that data size so hopefully you’ve got fast production & backup repositories to handle that throughput.

I’d also suggest at that core count to go physical proxy if you can, unless you really have a LOT of physical cpu cores in your virtual environment.

Also, for physical backups, you don’t need to size Proxies, there’s no such thing as a proxy server for physical server backups. The Veeam Agent is installed on the machine(s) you need to protect, and consumes resources on that device. It means you still need to consider sufficient resources on the protected server however.

well, maybe for doing a backup of physical server, it’s better for me to consider about transport mode, it using nbd or SAN

like you mentioned for physical server we don’t need proxy server for it.

but for VMs, the throughput from that formula, is it the throughput that has to be provided by backup repositories or is it the throughput that come out form our production site ?

Throughput is from production, as the proxy must read all data, before compression & deduplication and any other data efficiency techniques create a smaller backup. Normally you see a 2:1 reduction in data but that still means you should roughly have half the throughput speed available between your backup proxies to the backup repositories, and actually disk throughput on the backup repositories.

And yes transport mode will certainly impact proxy design choice. What is your production storage?

I use DAS for production storage, how many proxy that we need and what is the best practice for the scenario like image above ?

You cannot base Proxy design on a diagram and still need to use the calculation you used above.  When designing them best practice is 8vCPU and 16GB of RAM then scale out.  When designing them if virtual be sure to add extra SCSI controllers to them for attaching disks during hot-add mode as it can then process disks better.

As noted for your calculations a physical server would be optimal and if you can get access to the SAN where your servers reside you can use SAN integrated backups which are fast.

That BP site is a great one to follow but many things depend on your environment, backup window, SLA, RPO/RTO.

@Chris.Childerhose thanks for the advice, I will use that calculation for sizing the backup proxy need from now on 

This is a good thought….we tend to deploy multiple proxy’s when the need calls, but I hadn’t considered multiple SCSI controllers per proxy.  Is there anything special you do to use those controllers or just add them?

You just add them and when Veeam does hot-add it will use them automatically. 

Awesome...thanks for that info!

Not a problem. That is why we are here.