Target Repository: ReFS or NTFS

[bg.ranken]

So planning out our NAS backup with Veeam, we currently have about 25TB of files on our NetApp. Many of these files are expected to be duplicates because of our development environment, maybe about 5-10TB worth of the data.

So on that regards, what should we format the target repository as, ReFS or NTFS? From what I understand we wouldn't get any benefit from the ReFS for block cloning but we could get some benefit from built-in dedup on Windows with NTFS. None of our files are expected to be over the 4TB limit for Windows dedup as well. Looking at the data structure that Veeam uses it seems that the NTFS dedup would save us some space, but not really sure in the end. Veeam does have the compression built into the job itself, but it's also not clear if this is a global compression or just a per-file compression.

Regardless of space and dedup capabilities, are there any other benefits for using ReFS vs NTFS?

[Gostev]

In general, for our NAS backup NTFS or ReFS makes no difference. Also, built-in dedupe in Windows is a shared engine between NTFS and ReFS.
However, there's currently a bug on the Microsoft side in ReFS and Windows dedupe interop which impacts NAS backup specifically, so go with NTFS.

[bg.ranken]

Thanks Gostev,

Am I correct though in thinking that the global dedup will give us extra space savings? I know it might mean we have to set the Veeam compression to dedup friendly or none to get the maximum space from it, but am I correct to assume that the Veeam compression is not global?

[Gostev]

Yes, Windows dedupe will definitely add extra savings, but at the cost of reduced backup storage performance. I blogged about this in the weekly forum digest recently:

despite its limitations around max volume and max file sizes (and poor interoperability with ReFS block cloning) made it a bad candidate for image-level backup repositories, it is nevertheless a solid technology for many other use cases, for example NAS backup. We've been in touch for almost half a year now with one customer who was determined to use Windows deduplication for backup repositories used by file share backup jobs. Last we checked, they exceed 450TB worth of NAS data stored on a SOBR containing multiple 64TB extents backed by two HPE Apollo servers with 70 6TB hard drives, and are extremely happy how it performs – no issues whatsoever!