Advanced ReFS integration coming in Veeam Availability Suite 9.5

In my last post, I outlined a number of new enterprise scalability features: Backup acceleration technologies, full VM restore acceleration technologies and core engine enhancements. Part of the announcement for 9.5 was about the new advanced ReFS integration, which I’ll cover in this post as it is a significant technology for Veeam. Let’s dig in!

Advanced ReFS integration

The final 9.5 feature I will cover today is arguably the most amazing feature of 9.5: Advanced ReFS integration. Resilient File System (ReFS) is a new file system of Windows Server operating systems, and the Windows Server 2016 version brings new capabilities that Veeam Availability Suite 9.5 leverages in order to provide some truly incredible benefits.

The advanced ReFS integration supports ReFS volumes on internal, direct-attached storage (DAS) and Storage Spaces — both classic and Storage Spaces Direct (S2D). The benefits (explained below) will provide significantly faster synthetic full backup creation and transformation performance, as well as reduce storage requirements and improve reliability. Even more importantly, this functionality improves Availability of backup storage by significantly reducing its load — which results in improved backup and restore performance and enables customers to do much more with virtual labs running off of backup storage. This is a big deal, so let’s explore more ReFS integration.

Fast cloning technology

One of the hardest parts of making a backup, and in particular a synthetic full backup, is the transformation associated with moving blocks between different backups files in the incremental chain. Version 9.5 leverages the fast cloning capability in ReFS API to create synthetic backups without moving the data blocks between files, and instead, references backup file blocks already present on the volume. This means all manipulations associated with synthetic full backups are limited to metadata updates and require no actual I/O operations on backed up data.

The video below shows how the new advanced ReFS integration works compared to synthetic full backup transformations on general-purpose storage:

Below is a screenshot of the synthetic full backup being made on a VM backup on ReFS which took only 58 seconds to process three incremental backups. Compare this to over 36 minutes for the same VM backed up to an NTFS volume!

Advanced ReFS integration coming in Veeam Availability Suite 9.5

Spaceless full backup technology

Now that synthetic full backup creation and transformation are incredibly fast and not taxing on backup storage performance, you’ll probably want to have more full backups on disk — such as periodic synthetic full backups of primary backup jobs and GFS retention of Backup Copy jobs that can keep a set number of weekly, monthly, quarterly or annual backups. I have some good news for you here too: The spaceless full backup technology brought by the advanced ReFS integration allows multiple full backups on disk to share the same data blocks. Obviously, this means a significant reduction in overall storage capacity requirements for the backup repository, providing storage efficiency rivaling that of some deduplication storage systems, but without the performance impact associated with using those (data rehydration). And what’s best, all you need is a general purpose Windows Server 2016 with a bunch of internal or direct-attached disks — same as the backup server itself uses! How cool is that?

Check out this example of a ReFS repository with one synthetic full backup made:

Advanced ReFS integration coming in Veeam Availability Suite 9.5

Data integrity streams

While better performance and reduced disk space consumption is important, they are worth nothing without reliability, which is where the last advanced ReFS integration feature comes in to play. Version 9.5 will automatically enable ReFS data integrity streams on all backup files, which are used by background ReFS data integrity scanner to detect and notify you on silent data corruption. This is critical when it comes to ensuring a reliable forever-incremental data archival model for long-term retention. You can think of it as our existing storage-level corruption guard feature on steroids because it will keep verifying your entire archive, and not just the latest restore point.

Below you can see a screenshot of the Windows Event Log event from ReFS data integrity scanner catching an uncorrectable corruption of a file on simple ReFS volume:

Data Integrity Streams

All of that is available on simple ReFS volumes — but there’s even more for those using Storage Spaces, where its proactive error-correction functionality will automatically and seamlessly fix corrupted backup file blocks using a good mirror or parity copy of the affected block from another system, thus maintaining a healthy forever-incremental archive with the required level of redundancy.

As far as requirements go, they are rather simple: The backup repository must be installed on Windows Server 2016 (and most often — such as in the case of our backup appliances — it will be the same server as the backup server). This allows the repository to leverage the ReFS API that enables integration features identified above for greater performance and capacity efficiency with large-scale data sets, while ensuring data resiliency of Veeam backup data.

Do you agree that our advanced ReFS integration completely changes the game for Veeam users? Let us know in the comments!

Additional resources:

Get weekly blog updates
By subscribing, you are agreeing to have your personal information managed in accordance with the terms of Veeam’s Privacy Policy
Cheers for trusting us with the spot in your mailbox!
Now you’re less likely to miss what’s been brewing in our blog with this weekly digest.

Eliminate Data Loss
Eliminate Ransomware

#1 Backup and Recovery


  • Marc K says:

    This is really awesome.

  • andyr354 says:

    So this means ReFS has moved out of “Experimental” status for Server 2016 then? I know MS did not recommend using it to heavily yet in 2012R2.

    How does using ReFS in conjunction with storage spaces affect performance?

  • Rick Vanover says:

    Good observation, Andy. This a bit forward looking, in particular Windows Server 2016 being Generally Available and Veeam Backup & Replication 9.5 as well. At that point, it’s a support installation.

    In regards to the performance of S2D and data integrity streams; we’ve not been notified of any impact in that regard. However, we’re absent of large trials of real world usage with years of data ingested due to it being new…

  • James Davidson says:

    Will this work for Cloud Connect Backup Copy jobs if the SP is using Server 2016 with ReFS?

  • Anton Gostev says:


  • Anton Gostev says:

    ReFS is the recommended file system for Hyper-V 2016, enough said.

  • Neil Murphy says:

    Very exciting stuff. I’ve been using Timothy Dewin’s Restore Point Simulator for sizing repositories ( Do you know if Timothy plans to update it for ReFS?

  • Rick Vanover says:

    Good question Neil. I doubt it for now, as we need some “run time” of history to collect some best practices first.

  • Rick Vanover says:

    And that is awesome.

  • jeerd says:

    Do the KB1745 Deduplication Appliance Best Practices apply to ReFS? In particular the Advanced Settings -> Storage?

  • Hi,
    No they do not because ReFS is not deduplication. In fact, because of the way the technology works (and the API calls we do) you will want to use synthetic fulls

  • Jacob Benson says:

    This just made my day

  • starforce1 says:

    can you illustrate how a 14 days retention period work? What will happend to the full backup bk1 when we get to day 15?

  • starforce1 says:

    do you have to use ReFS on the source side?

  • Rick Vanover says:


  • Rick Vanover says:

    Nothing specifically different would happen (based on your configuration – forward/reverse incremental chain). Just that when it does need to move data, it can do so in accelerated fashion with ReFS.

  • Darren Toews says:

    Can you combine deduplication with ReFS? Either with Microsoft’s own volume based dedupe, or with a virtual appliance like HPE’s StoreOnce VSA? Or is it a case of diminishing returns (ie: actually using up more space as in compressing an already compressed file)?

  • Rick Vanover says:

    No. There are storage efficiencies for ReFS however, it’s not deduplication per se.

    As for an appliance or VSA, it’s just going to burn CPU needlessly for low gains. May be worth trying as everyone’s data is different – but this is not a scenario we are recommending.

  • Darren Toews says:

    Good to know. Thanks Rick!

  • Magnus von Rosen says:

    Are there any particular settings to optimize when storing on a local ReFS volume, in Veeam? Do I let Veeam stay on the defaults? Deduplication = on and compression = default etc?


  • Rick Vanover says:

    The best part Magnus is there is no configuration / setting! It just detects WS2016 and ReFS and you are good to go. The only catch is IF it is Windows Server 2016, and you are using NTFS – it will warn you that you are not doing it in the best configuration – so you only get warned if you do it wrong :)!

  • Magnus von Rosen says:

    Right! I noticed that. I borrowed a customers server, doing a test comparison right now. Small Hyper-V enironment, 3 VM’s, One job running to NTFS with dedupe (no dedupe in veeam), one job running to ReFS with Veeam dedupe on.
    I have done a series of reverse inc and active Full runs now, inserting various data into the VM’s as I go.
    In the ReFS Job, I can see the VMs beeing backed with the [FastClone] tag. Very fast.
    I’m just not sure how ReFS saves me any space? Each “Active Full” run is still an extra 33GB each. Right now NTFS is using 66GB and ReFS 100GB, after 5 incs and three Active Fulls.
    I was sort of expecting the Active fulls to be very small. What did I misunderstand?

  • Rick Vanover says:

    it will save you space when you do any synthetic full backup OR GFS retention points.

    This storage efficiency is based on the referenced blocks. I’m working on a whitepaper on this just now :)

  • Duncan Crothers says:

    Hi Rick,
    Great article! Do you have a link to the whitepaper? Is it Veeam’s position that ReFS on Windows 2016 is best practice?

  • Rick Vanover says:

    The paper doesn’t seem to be published yet – it will be soon at the page..

  • Anguel says:

    Is this usable at all on a single Windows Server 2016 with a storage spaces classic mirror and ReFS? I read everywhere that it becomes very slow and unusable because integrity streams are turned on.

  • Rick Vanover says:

    If it’s a single server – I’m not sure classic mirror would work. I’d not do the mirror with the single server myself.

  • Anguel says:

    Thanks a lot for the fast reply, Rick. Do you mean it will not work at all or will it be slow?

Leave a Reply

Your email address will not be published.