Tasks for Repos are based on CPU cores versus RAM as per - Backup Repositories - Veeam Backup & Replication Best Practice Guide

So it is indicated to start with one and the scale up but not have too many which burdens the disks.

Depending on OS as well - NTFS, ReFS you will need to change the numbers.

Also keep in mind the OS uses cores for CPU too so it is a balance.

From the site above -
 

Concurrent Tasks

Start with configuring one concurrent task per CPU core and adjust based on load of the server, storage and network.

Dependent on the used storage too many write threads to a stroage might be counter productive. For example, a low range NAS storage will probably not react very well to a high amount of parallel write processes created by per-VM backup files. To mitigate these effects its better to limit the concurrent tasks in this case. On the other hand a high-end deduplication appliance might have a limit on a single write thread but can handle a lot of parallel tasks very well and thus profits from the use of per-VM backup files and enough concurrent repository tasks.

** Note: Consider tasks for read operations on backup repositories (like backup copy jobs). **

Hi @BirdMan ...sorry for the vague answer, but “it depends”. My recommendation, and what I would do, is leave the max concurrent for each repo. You could then stagger your jobs such that they would not use the same Host the Repos are on at the same time and your jobs would “rock out” on the Host resources. I actually do that myself. And, Veeam would automatically ‘hold back’ on tasks anyway if they are “in use” by other tasks/jobs.

Hope that helps.

Cheers!

What I also recommend is viewing Tim Smith’s “Architect Sizing” Session from VeeamON. It’s pretty amazing and would no doubt help you in moving forward. Go to veeam.com/veeamon and ‘log in’, then view Sessions from the Americas. Video is here.

Cheers!

Yes this is a great way to see how to size things also.

Also see this site - Repository Design - Veeam Backup & Replication Best Practice Guide

Thanks for the quick answers. So in the case of cpu we have 32 Cores and just increased from 32 to 64GB RAM running REFS. I had limits on my concurrent tasks at 6 in each repository. From what I am hearing here we should incrementally increase the tasks if desired, while monitoring RAM CPU and Disk IO.

That to me would be the recommended approach to see where the sweet spot is.  You could do 32 tasks but then you are hindering Windows also.  It is a balance between tasks and how many VMs/VMDKs you are backing up.