ReFS success stories?

[ivordillen]

Hello,
I recently updates to 9.5 and am putting my first jobs in ReFS. The ReFS is on a second environment but at some point you have to go all the way. The advantages of ReFS are huge so I'd like te push it but...

Are there a lot of success stories? It is maybe not important to say how much improvements you see but how much volume you have in ReFS.

Thanks a lot.

Ivor

[Gostev]

What is your backup volume size? We've been seeing good feedback from users with 64KB cluster size ReFS volumes of tens of TB in size. There's currently not enough data yet for 100TB+ volumes, as most larger customers usually wait for at least half a year before starting to adopt new features and technologies.

On the other hand, ReFS volumes with 4KB cluster size (default!) seem to have all sort of issues on larger volumes at this time, so avoid those for now. Although judging on the recent Windows Server 2016 updates, Microsoft is aware and working through those issues.

[dgomes]

Many of my customers have gone from 6 - 8 hour synthetic fulls to about 20 minutes on 64KB ReFS.
Nothing but success on my end. The only real issue we are having is customers who just recently purchased 2012 not wanting to switch to 2016 to take advantage of this, it's very annoying from an infrastructure design point of view.

[rendest]

Absolutely avoid 4K cluster size at all cost, it is still unstable, even after the patch. Our backup repo with 64G memory and two quad core CPU couldn't cope with a 64TB ReFS volume (first 4K, later on 64K).

What we are currently seeing is a huge performance drop & instability issues after volume sizes of 32TB (filesystem becomes unresponsive). And Microsoft even mentioned not to go over 64TB, period (veeam-backup-replication-f2/refs-4k-hor ... ml#p227546)

[Gostev]

Microsoft even mentioned not to go over 64TB, period

More like one specific support engineer did. I am checking with the ReFS devs, and pretty sure they will disagree at least in general - I am not ruling out the possibility of some existing teething issues considering Server 2016 age.

[DaveWatkins]

We went from RAID10 to RAID6 LUN's when converting to ReFS with 64k clusters and our backups are still faster (with reverse incremental even). We did also go to a larger stripe size but so far the 64kb cluster sized LUN's have been fantastic

As mentioned, do not use 4k cluster sizes when formatting ReFS