ReFS, Fast Cloning and Reverse Incremental B2D2T jobs?

[cerberus]

Hello,

I read on this veeam-backup-replication-f2/refs-fast-c ... 44079.html that it is not suggested to have fast cloning for jobs that use reverse incremental backup type, is this true? Thread is a few years old.

We are not doing any active full backups since we are running in reverse incremental so last backup is always a full backup.

We need to do reverse incremental as this is (as far as I know) the only way to get full backups to tape every day which is a requirement.

If we do incremental backup, only the incremental data will be offloaded to tape.

In a ReFS formatted repository, doing reverse incremental daily backups of VMware VMs (CBT in san mode) where does Fast Cloning come into play for speed increase?

[Gostev]

Hello!

Yes, it's true. Fast cloning significantly reduces I/O load on the backup repository, removing read+write operation of moving replaced data from VBK from VRB. So with fast cloning, reversed incremental backup requires 1 I/O per block (write VBK) instead of 3 I/O per block without (read VBK + write VRB +write VBK).

Thanks!

[cerberus]

Gostev,

So the best practice from the linked thread to "not use reverse incremental with ReFS due to the backup mode causing additional fragmentation" - we we can disregard that part and use ReFS with reverse incremental backups to gain speed...? Should I worry about fragmentation at all?

[Gostev]

Basically, you can get faster incremental backups at the cost of slower full VM restore. Granular and instant VM recovery performance should be largely the same, since these are random I/O in any case.