Replication Maximums and Consumption

[LittleNickey]

I'm looking into using Veeam Replication (not CDP Replication) and trying to find information regarding some maximums and how to calculate resource consumption, but can't really find anything in the documentation regarding this.

Earlier this year I took an official VBR course, and from my notes the instructor said something in the lines of this:

1 data mover per source VM + 1 data mover per target VM = 2 movers/replicated VM = 2 cores/replicated VM

Is this correct and would this mean that the backup proxy at the source location uses one CPU core per VM and target backup proxy as well respectively, or if source and target is the same proxy 2 CPU cores, per VM replicated? So a proxy with 10 Cores (for easy math) would be able to replicate 10 VMs at the same time, or only 5 if it is both source and target? Is this also true no matter which replication type used (seeding, from production storage, from backup files, from storage snapshots)?

And what are the Min and Max RPO times supported for Veeam Replication? I can only find information regarding CDP replication (2s/15s-60m). In the schedule option you can set "Continously", would that mean RPO 0 (Zero)? I did however read somewhere that CDP replication allows "better"/lower RPO's, so I'm guessing there is some limits to Veeam Replication.

Primarily I'd like to know how many VM's I would be able to replicate, and if that is simply related to the hardware assigned to the proxies then how to calculate it. If it is related to number of cores, what happens if that number is exceeded? Will it fail, take longer time, or would I manually need to split them from running too many at the same time etc?

Thanks

[PetrM]

Hello,

I guess the best explanation of our resource scheduler logic is given on this page, please review it. We recommend sticking with 2 concurrent tasks per 1 CPU core. Every concurrent task is created per each VM disk. If the same machine has the roles of source and target proxies or the roles of repository and target proxy in case of seeding, you need to make sure that there are enough resources to process the cumulative number of tasks.

For example, you need to replicate 4 VMs, every VM has 2 disks. If you want to process all of these VMs in parallel, you need 8 tasks on the source proxy and 8 tasks on the target one. If the same machine has both roles, you need to set 16 tasks. Since we recommend running 2 tasks per CPU core, 8 cores is enough in this case.

If the number of cores is exceeded, it's difficult to predict what kind of issues you may face: it could be just longer runs or intermittent failures. I would not recommend exceeding the recommended number of cores, otherwise, we cannot guarantee stable functioning. Instead, it's better to deploy more proxies, for example, virtual ones to process VMs in HotAdd mode.

There is no strictly defined minimal or maximal supported RPO, it's up to you how to schedule jobs. The continuous mode does not mean RPO 0, it's just a type of job schedule. In continuous mode, a job is constantly running but we don't know how fast will it process the workloads. Obviously, I'd expect a lower RPO with the continuous schedule than with the daily schedule. But if you're seeking the minimal possible RPO, CDP is a way to go as it provides near-zero RPO.

Thanks!

[LittleNickey]

Thanks @PetrM!

So a Veeam Proxy with 20 cores would then be able to protect 20 VM Disks concurrently if it is both source and target, and 2 proxies (one source and one target) with 20 cores each would be able to protect 40 VM Disks concurrently. So you could schedule multiple jobs in intervals, depending on how long each job takes of course, to protect multiple groups of VMs with a max of 20/40 disks? Is that the gist of it?

And a follow-up question: Would the Veeam Server be consuming any major resources as well (i.e. 1 core/2 tasks) or would it only be contained in the proxies?

[PetrM]

Hello,

Is that the gist of it?

Your understanding is absolutely correct.

Would the Veeam Server be consuming any major resources as well (i.e. 1 core/2 tasks) or would it only be contained in the proxies?

Tasks and cores are purely about proxy server resources because these components are responsible for data processing. Veeam B&R server just manages running jobs, sends commands to different backup infrastructure components, updates job statistics and information in the configuration database. You should have 4 GB RAM minimum + 500 MB for each concurrent job and we recommend having 4 cores minimum to maintain the management operations.

Thanks!

[LittleNickey]

@PetrM, thanks for that!