Job Scheduling with "Parallel Processing" enabled

[pinkerton]

Dear Forums,

we've started using Veeam with version 6 back in 2011 and have not used parallel processing since then. We are now moving our Veeam Server from Windows Server 2008R2 and Veeam 8.0 to a new physical Server with Veeam 9.5 and Windows Server 2016. Within the scope of this migration we are also looking into using the "Parallel Processing" feature. I've already read the corresponding parts of the best practice and user guide but still have some questions regarding this topic. Let's assume the following:

- One Physical Server with "Parallel Processing" globally enabled
- Only one backup proxy is used with 8 concurrent tasks enabled
- Only one repository is used with 8 concurrent tasks enabled

Is it true that:

- ... if I start a Backup Job with 20 VMs all 8 tasks are used in parallel? Let's say: The first and second VM in the job contain 5 hard disks each. Are all 5 disks from the first VM + the first 3 from the second VM being backed up in parallel?

- ... another Job cannot start until there are free task slots available again? Let's say: The Job with 20 VMs is running and at 03:00am all 8 tasks are being used for VMs within this job. What happens if a second job is scheduled to start at 03:00am? Will it be put on hold until free Tasks slots are available again? Which means it could start a lot later than intended I guess?

Let's say it is mandatory that the second job starts at exactly 03:00am: How could this be solved? Is this even possible at all with parallel processing enabled? I guess I could just configure a dedicated Backup Proxy for the Job, which however requires a second server and is no option in my case.

I however can create multiple repositories pointing to the same storage. Could it be a workaround to point Job 1 to a repository configured with only 4 tasks (so 4 from the 8 tasks of the proxy remain unused)? Then I could point the second job to another repository that could use the 4 remaining tasks from the proxy and it is made sure that it could start exactly at the scheduled time.

Thanks
Michael

[PTide]

Hi,

... if I start a Backup Job with 20 VMs all 8 tasks are used in parallel? Let's say: The first and second VM in the job contain 5 hard disks each. Are all 5 disks from the first VM + the first 3 from the second VM being backed up in parallel?

Yes.

... another Job cannot start until there are free task slots available again?

Yes.

What happens if a second job is scheduled to start at 03:00am? Will it be put on hold until free Tasks slots are available again? Which means it could start a lot later than intended I guess?

Yes, it will be put on hold until Job1 starts its last disks and there are free slots left. Yes, that might delay the start of the 2nd job.

How could this be solved?

A) Add a separate proxy
B) Increase the amount of slots
C) Use the backup window
D) Use the "multiple repo" approach that you've described (it will cause the proxy to underperform though)

Thanks

[pinkerton]

Thanks, that helped me a lot!