Groupings of Users for Jobs

[EricinIT]

In previous versions, I've been told to use M365 groups to split up jobs for large organizations. Now with Proxy Pools in place, is using M365 groups still the best approach to run multiple backup jobs for large organizations. Can we use one backup job to do the whole organization? I'm looking for recommendations for backing up a tenant with 10,000+ users.

Thanks,

Eric

[Mildur]

Hi Eric

I have reached out to QA to clarify maximum users/objects (mailbox, archive, OneDrive, personal Sharepoint site) per job.
Stay tuned for an update.

Best,
Fabian

[EricinIT]

Thank you!

Eric

[Mildur]

Hi Eric

I'm still waiting for a definitive answer. But Job splitting will still be important.

I'm currently verifying the maximum number of objects per job, which our recommendation is still not more than 40000 objects per job.
We need to determine whether we should create four jobs with 10,000 users each (protecting all workloads) or four jobs that include the entire organization but only protect selected workloads:

- 1 Job (Entire Org) = 40,000 Mailboxes
- 1 Job (Entire Org) = 40,000 Archives
- 1 Job (Entire Org) = 40,000 OneDrive
- 1 Job (Entire Org) = 40,000 Personal SharePoint Sites

We want to ensure that enumerating all 40,000 users multiple times doesn't overwhelm the Graph API or cause throttling issues with Microsoft's endpoints.

Best regards,
Fabian

[EricinIT]

Ok, that makes sense.

Eric

[Mildur]

Hi Eric,

Thank you for your patience. Our QA has confirmed that we can split backup jobs as I suggested.

By splitting the jobs per workload type, you will be able to add the entire organization to the job, ensuring that new users in Microsoft 365 are automatically protected by your backup server. There's no need to use M365 groups or scripts to manage users per job.
For small tenants, I would personally create a single job if all workloads have the same retention setting. But for larger tenants, like the one in your example, it's best to split the jobs by workload type.

Please note that the current recommendation is to limit each job to 40,000 objects based on our testing. However, if you exceed this number by a few hundred in a single job, that's still acceptable.



Best,
Fabian

[EricinIT]

Awesome, thanks for checking into it and letting me know!

Eric