Different backup throughput speed depending on VM size

[GromPit]

Hello,

I am running PVE cluster (3 nodes, all version 8.4.5) and PBS in version 3.4.3. As a storage I have two NFS QNAPs with 24 NVMe disks each - one of them is being used for hosting VMs, the other one for PBS as backup storage. Everything is working fine and stable, although I've encountered very strange (for me) behavior while creating backups with PBS. It seems to me, that throughput speed (not backup time!!!) relies on VM Size - the bigger the VM is, the slower the throughput is.

For instance (all initial backups)

-backup of 2TB VM took a little over 1 hour (backup was sparse, 50%)
-backup of 1.2TB machine took 50 minutes (backup was sparse, 32%)
-backups of small machines 100-300GB took up to 5 minutes (sparse).

I wonder is it normal and how can I verify the actual backup speed (backup status doesn't tell anything precise - it increments by 1%). For example, now I am making initial backup of huge 8TB VM and after half an hour I backed up only 5% which is a little bit over 400GB.

It is all strange to me. It seems that there is no connectivity issues. Both NAS are connected via 4x25Gbit to switch - the servers are connected via 2x25GBit. NAS is on different subnet than management.

Would anyone be so kind to shed some light on my issue?

Best regards,

Piotr

 

[SInisterPisces]

PBS deduplicates when backing up, so there's a lot going on aside from just network throughput. Depending on what options you have set, the backup job has to, in no particular order (1) compare the data; (2) break the data down into chunks; (3) apply deduplication to reduce duplicate chunks; (4) perform compression; (5) perform encryption; (6) possibly other things.

The amount of time it takes to perform all those operations will vary by VM and LXC. The bigger the guest is, the longer it will take to do the computational operations; network or storage I/O performance doesn't impact those directly.