Is ReFS safe to use? Should I be using it?

[CloudMSP]

1. I have one cloud connect server/repository, and another repository both with the specs below

R710 - 32 GB - H700 - 18TB RAID 10

Should I be using ReFS on these ?

2. My BDRS on site are Dell T30 with 12 TB of Windows Storage Pools (SATA)

Should I be using ReFS on these?

I've read about major problems and it's got me scared, but also good benefits. Is ReFS safe and recommended for use in these two cases?

Is it worth the effort to switch them up? Or just do it as they are replaced?

The largest backups are about 5TB.

Thanks.

[Mike Resseler]

Hi CloudMSP,

You are absolutely right in saying that the issues that you read about ReFS should scare you. But there is good news. You might have read the word from Gostev this weekend and saw that the necessary patch for these problems is available (again). https://support.microsoft.com/en-us/hel ... -kb4088787

There is still an issue with memory consumption and they are working on a patch for that but this is not the case for everyone and is more for extremely large repositories. We do think you should look into formatting the ReFS volume with 64k instead of 4k.

For your BDRS on site, will there be synthetic full backups? In that case it is certainly worth it in investigating and formatting with ReFS and see the benefits. The same requirement obviously stands from above. The biggest benefit is that block cloning technology that we use. Merges will be much faster and even use less storage because of the linking and not actually copying data anymore.

Considering the both sizes (18TB and 12TB) I wouldn't worry too much and with the patch it is safe to use.

Hope it helps
Mike

[CloudMSP]

Thanks!

The BDR's are per VM files in forever incremental mode, does it matter then?

[Mike Resseler]

ReFS will only give you benefits when you do synthetic fulls on a regular basis. So if you have forever incremental mode, then there is nothing to gain with ReFS

[CloudMSP]

ReFS will only give you benefits when you do synthetic fulls on a regular basis. So if you have forever incremental mode, then there is nothing to gain with ReFS

What about merges, improved?

and on the CC repository side?

Actually what is the benefit of a synthetic full if it's just pointers to the same data? The data that you are fearing to go bad, which is the reason you made the full in the first place?

[Mike Resseler]

Correct on the safety. That is indeed an issue that can arise, but ReFS comes with a functionality that it can detect issues with your data. Now when it is just a simple server, then it can't FIX it (that is why ReFS is mostly used on primary repositories and there is still a backup copy job to other locations or media). If it is ReFS with S2D or classic storage spaces, then it can provide healing capabilities

Makes sense?

[jmmarton]

What about merges, improved?

Correct. Even when using forever forward incremental there is still a performance improvement for the merge/transform process when using ReFS.

Joe

[CloudMSP]

Thanks guys I'm making the switch where ever possible and moving forward.