New Backup Repository Install. ReFS or NTFS?

[kevin.boddy]

Hi,

We are busy installing a new Windows 2016 Super Micro storage server backup repository.

2x 8 core CPU's
256GB Memory
60x 10TB hard disks
Hardware RAID controller

After the ReFS back-port patches that were released the previous week, should I be looking at using ReFS again on this system?
I was advised by Veeam 2 months ago to stick to NTFS but alot has changed since then.

I know the 1GB to 1TB rule for memory isn't being followed here but Anton did mention that past 128GB it's non linear, I'm not sure if the 256GB is ok or not.
What I'm trying to solve with my current storage, which is currently incremental forever, is to get away from the merge opertions which are IO intensive.
My previous experience with ReFS last year was that it worked great and then suddenly became unusable after a couple of months and used to just stall the entire system.

Any advise now, ReFS or NTFS?

Thanks
Kevin

[JHuston]

Honestly for me stability trumps all. I'm also in the stages for planning new servers for repositories and will be going NTFS. I just don't have time to risk my repo's going down, loosing backups, having to redo the repositories and re-running backup jobs, performance issues, patching issues etc.

ReFS sounds great and all but to me it's still just too risky.

[ejenner]

Could go 50/50 for a while and see which you're more happy with. That would halve the amount of work you'd have to repeat if you lost a repository.

After a few months delete the volume you don't want and extend the one you're keeping. Or maybe add a new one of the same kind. Better not to create humongous volumes anyway as it's more trouble if things do go wrong.

[Mgamerz]

The new ReFS updates have fixed the stalling issue, at least in part. I haven't had a stall in about 2 months now. The forward merge on ReFS is like 40 mins for 12TB instead of like 100 hours.

[kevin.boddy]

Hi

Thanks for the suggestions. Think I will try a 64TB ReFS volume and see how it goes with the latest patches.

Thanks
Kevin

[lando_uk]

If you don't mind me asking, how much did that supermicro beast cost you?

[kevin.boddy]

If I convert back to USD just under 33000.

[Steve-nIP]

Honestly for me stability trumps all. I'm also in the stages for planning new servers for repositories and will be going NTFS. I just don't have time to risk my repo's going down, loosing backups, having to redo the repositories and re-running backup jobs, performance issues, patching issues etc.

ReFS sounds great and all but to me it's still just too risky.

That's also the way I feel about it.

[andyrbates]

We experienced a lot of issues with ReFS which resulted in us losing data on backup jobs. We have reverted to using NTFS for Backup Jobs and ReFS for Backup Copy Jobs and have not experienced any issues. We will consider moving the Backup Repositories to ReFS when we are confident the issues have been resolved. We have found space-saving benefits of 25% using ReFS.

[csydas]

Personally I'd get it on a stable version of ReFS and then just make darn sure everyone in the shop knows what's up and not touch it until the brave souls here on the forums test it out first. Our primary repos are all a mix of ZFS and BTRFS (love love love love volume snapshots), but we also have a couple of ReFS ones up for a few sites, and I can't deny that when ReFS is stable, it's awesome, and it blows our other backup times out of the water completely.

[evilaedmin]

Our primary repos are all a mix of ZFS and BTRFS (love love love love volume snapshots), but we also have a couple of ReFS ones up for a few sites, and I can't deny that when ReFS is stable, it's awesome, and it blows our other backup times out of the water completely.

What OS/version do you use for ZFS? Are you using it for volume management in place of hardware raid as well?