What are the Advantages of PBS vs PVE Backups?

[+DS_DV+]

Hi,
sorry if this question got answered before :/
I currently fail to understand why i should install PBS on my PVE since i allready have a Backup job running which is storing the snapshots on my nas.
Can someone explain to my what advantages it would bring to now apt install proxmox-backup-server on my node and setting up the nas share inside the PBS instead?

I understood that most folks run PBS as separate machine which can also do storage management.
But for the sake of optimizing footprint and storage management i would like to keep PBS (if used) on the same machine as PVE
(the target storage [NAS] is backed up)

with kind regards and thank you for your kind patience

 

[UdoB]

...which is storing the snapshots on my nas.

As Vzdump-Files? If yes: each and every backup uses the same space.

A small Odroid with an SSD datastore and just ~2 TB occupied for PBS tells me:

CT 7 Groups, 109 Snapshots
Host 6 Groups, 13 Snapshots
VM 123 Groups, 2649 Snapshots
Deduplication Factor 37.24

I am storing 2649 backups with a native (compressed) size of ~75 TB here. De-duplication is magic

(This is a homelab; more than 98 % Linux. The dedup-factor is so high because half of the VMs are turned off most of the time --> zero modified data...)

 

[+DS_DV+]

Ohh so PBS offers deduplication as a service if your storage solution does not do it on its own
that is indeed a nice feature to have.

 

[edwin2024]

And you can restore a single file. As far as I understand pce backup you have to restore an entire vm at once. No single file selection.

 

[Dunuin]

There is live restore where you are able to start and use a VM that isn't restored yet.
There is encryption with zero-trust in mind.
There are backup integrity checks.
There is build-it sync jobs to replicate your backups between different PBS (offsite backups etc).
In contrast to VZDump there is ransomware protection possible as you can set up privileges so that PVE is only able to create or restore backups but it isn't allowed to overwrite or delete any backups. All retention then could be handled by the PBS itself.
...

But running PBS on your PVE node isn't great:
- you will lose that ransomware protection
- unlike when using VZDump, there won'T be any backup archives/images you could directly restore to your PVE from your NAS. Without a running PBS all those backups on the NAS will be useless.
- PBS needs massive IOPS performance (thats the reason why you should use SSDs and not HDDs to store your backups) and when using SMB/NFS the network latency will hurt that IOPS performance

So best you got two dedicated PBS servers. One onsite and another one offsite pulling backups over VPN from the onsite PBS.

 

[+DS_DV+]

ok i will install PBS

@Dunuin do you know if it is possible to run the backups on local storage and after they are finished sync them to a NAS ?
To get the performance but without the need to run extra local storage?

 

[Dunuin]

ok i will install PBS

@Dunuin do you know if it is possible to run the backups on local storage and after they are finished sync them to a NAS ?
To get the performance but without the need to run extra local storage?

Yes, but the stuff that needs performance are the maintainance tasks like GC and re-verify. You still need very fast storage on your NAS as you would need to run those there too.
And again, without a working PBS you won't be able to restore any backups from your NAS.
So best to have a dedicated server with local storage just for PBS.
Or at least I would run that PBS as a VM directly on your NAS in case your NAS is capable to hosting VMs.
In case you virtualize your PBS in a VM on PVE make sure to store VZDump backups of that PBS VM on some NFS/SMB shares on that NAS so you could restore PBS without a PBS running to be able to restore everything else from PBS.
But without SSDs its a pain to run PBS in case you plan to store several TBs of backups. Those tasks will run for hours or even days and your NAS won't be usable as the HDDs will be 100% utilized while doing it. There will be millions over millions of random IO.

 

[+DS_DV+]

Thank you for the clarification

Currently i have a Seagate Exos X - X16 16TB ST16000NM001G 3,5" 256MB Cache HDD in there.
But i could add a Sata SSD as a Second Storage pool.
Or maybe i could add an USB 3.0 SSD directly for PBS.