ReFS and backup mode

[tom.starren]

Hello,

Next year we are going to upgrade our storage and proxy servers to Windows Server 2016 and ReFS for backups. One thing I'm not sure which backup mode to use on the primary jobs. Reversed incrementals could be viable with ReFS as is Forever incremental, but is it safe to only use synthetic backups? Currently I want to use Forward incremental with scale-out repo (incrementals on ReFS volume and fulls on NTFS volume with deduplication enabled).

Thank you for your advice.

Kind regards,
Tom

[jmmarton]

The synthetic fulls on ReFS is where you'll really see the space savings and the IO savings (during merge). Is there a reason why you want to put the fulls on NTFS instead?

Joe

[DaveWatkins]

To give some real world experience.

Just before christmas I completed our migration. We went from a 40 disk (7200RPM) RAID10 array with 2 LUN's presented to the repo server, both formatted NTFS with 64k clusters on 256k stripe size, over to 2 x 20 disk RAID6 arrays with a LUN from each array presented to the repo server. Those were formatted ReFS with 64k clusters on 512k stripe size.

We've always used reverse incremental and our backups are now significantly faster. It's possible our conversion to RAID6 and the I/O savings from ReFS end up cancelling each other out and our increased performance is just the improvements in 9.5 as I didn't have a lot of data from after 9.5 was installed but before the ReFS volumes were made available but overall it looks like we've seen improvements in speed from the disk changes too. They might also be coming from the larger stripe size but at the very least we didn't lose anything from the RAID6 conversion and ReFS

[tom.starren]

The synthetic fulls on ReFS is where you'll really see the space savings and the IO savings (during merge). Is there a reason why you want to put the fulls on NTFS instead?

Joe

is this a "safe" way of storing your primary backups? In the past I thought active fulls where advised to ensure that the data is readable. Would you then use reversed or forward incremental (RAID set will be RAID6).

[foggy]

With SureBackup and built-in health checks in-place the need for active full backups is not that great.

[tom.starren]

Hi Alexander,

Surebackup isn't setup and wasn't on the planning to do. We will start using repositories which reside on ReFS (Windows 2016, no S2D). Would you advice then to run an active full after x time? Will use a retention of 7~14 points (daily backup) on primary storage and a GFS structure on the two secondary storages.

[foggy]

Yes, you can schedule periodic active fulls to avoid possible issues incurred by extremely long backup chains.