ReFS Block Clone Synth Full Space Saving - Per Machine Backup Files Makes a Difference?

[rleon]

Hi all,

Just want to know if enabling "Use per machine backup files" mode in the repository makes any difference to the ReFS Block Clone (Veeam Fast Clone) Synthetic Full disk space savings.

For example, I have 2 repositories, both with ReFS 64K Fast Clone enabled and working.
One of them has the "Use per machine backup files" enabled, the other with it disabled.

If I backup 10 VMs daily, with Saturdays running synth full, then after a month, would using one repository over the other result in better ReFS Block Clone disk saving amount?

Thanks!

[Mildur]

FastClone reuses already existing blocks from the backup chain on the repo for the next synthetic fullbackup. Both methods are ok with Fastclone.

„Use per backup Job Backup Chain“ will have better storage optimization because the deduplication is global in the job for all vms inside.

„Use per machine Backup chain“ will work too with FastClone, but it will uses more space on your storage because the deduplication is not global for all vm inside the job.

I recommend to use „Use per machine Backup chain“ if you can. If you have a an issue with restore points from a vm and need todo an active full, only one vm in the job needs that and not all of them.

[rleon]

Ah yes, the job-scope-dedup savings should be factored in as well. Good to know FastClone is still effective. Thanks.

[mkretzer]

One other important thing: Since you can't "migrate" the block cloned backups without "inflating" them using "per backup Job Backup Chain" means that when one of your repos in a scale out backup repo runns full the backups of all machines must move to the other extend. With per-VM you only loose block cloning of one or two VMs in that situation.
Also, at least in V10 per-VM was faster because it could use multiple IO streams. I don't know if this is still relevant in V11.

[Gostev]

Yes, this will always be relevant... enterprise storage "likes" multiple streams. I recommend using per-VM for all environments with more than a few dozen machines due to performance and management benefits. Besides, we're enhancing per-VM jobs massively in V12 with new management capabilities which are simply impossible when all VMs go into the same backup file.