Data deduplication in Windows 10 and Windows 11 using Microsoft tools

The situation changed with the release of Windows Server 2012 – now in any server OS, deduplication has become part of the file server role. The functionality was developing well, not only in large LTSC server builds, but also in the Semi-annual publications published every six months, receiving new features.

In this article, I will not dig deeply into the architecture of the packages from which the Windows OS is assembled, but I think that I will not tell you a secret by saying that client OSes differ from server OSes essentially only slightly. Yes, there is a lot of “server” stuff there, or vice versa, they don’t have a lot of “client” stuff (for example, Bluetooth or IrDA), but in general they grow from the same root. And in those happy years when the Windows Server and client Windows builds are based on the same kernel, we can easily borrow some of the functionality from the server to the client, or vice versa.

About assemblies and versions. Microsoft now usually calls a version a four-digit number consisting of the last two digits of the year and two digits of the month. For example, the version of Windows 10 May 2019 Update was 1903, and the build number began with the numbers 10.0.18362 (or 18362 for short), it was paired with Windows Server 1903 (aka 19H1). With a simple update, they were updated to Windows 10 November 2019 Update and Windows Server 1909 (19H2). Then, with the advent of Windows 10 May 2020 Update (and Windows Server version 2004 or 20H1) and until the end of support for Windows 10, the reign of builds 10.0.1904x (19041, 19042, 19043, 19044 and 19045) reigned. Simple updates brought any of these OSs up to date. And if you applied any components from the server OS to the client OS, they were updated along with the client OS (which will be discussed a little later).

With the release of Windows 11 in October 2021 (version 21H2 build 10.0.22000), the situation worsened somewhat – there were no versions of Windows Server on this kernel. Just as there were none based on 22621/22631 cores for Windows 11 2022/2023 Update.

Now, fortunately, the time has come again for the long build 10.0.26100, on which Windows 11 24H2 and Windows Server 2025 are based, and it is already clear that the next edition of Windows 11 in 2025 will only be an incremental update – that is, compatible with files Windows Server 2025, which will also receive Server Core updates as part of the Annual releases.

So, enough of the lyrics, let's get down to practice. I will not go into detail about how exactly packages are created for portability between OSes – if there is interest, write in the comments, perhaps this is worth a separate article. But I will give a link to SxSv1 utility, which allows you to export components from any installed or mounted (WIM) OS. Below I give examples of packages that I exported from evaluation images of Windows Server 1903/19H1, 2004/20H1 and 2025/24H1 using this utility. You can easily do the same – there is no piracy involved. Get a working but unsupported solution – it's up to you to decide what to do with it. Yes, it will receive feature updates on Windows 11 24H1 through Windows Update if such updates are included in the cumulative update – which Windows 11 and Windows Server 2025 share.

Actually links. Deduplication files for Windows 10 versions 1903 and 1909 (builds 10.0.18362 and 10.0.18363). Deduplication files Windows 10 2004/21H2/22H2 (builds 19041, 19042, 19043, 19044 and 19045). Windows 11 24H2 Deduplication Files (builds 26100/26120).

Downloaded the archive, unpacked it, launched the CMD file from the administrator, rebooted, use it. You now have the required data deduplication in your components.

Using the ddpeval utility, you can evaluate the benefits of using data deduplication:

In my case, all real disks have long been deduplicated, and I am doing a demonstration using the example of a VHDX disk of one of the VMs. There I see that out of 15.57GB, when deduplication is enabled, I can free up 6.19GB. Enabling deduplication is simple:

I do not enter optional parameters; they do not seriously affect the result. After some time (leave it overnight or run optimization scripts manually), you will be able to see the result:

In my case, on a 1 TB disk, 380GB are actually occupied, and another 486GB are optimized by deduplication. I have virtual machines on this disk. On the disk with backups of a home computer, the statistics are even more impressive.

I’ll end the story here, because this is essentially just advice to enthusiasts – don’t be afraid to try new functionality, including from server OS to client OS. As a fan of virtualization, for example, in addition to deduplication, I also transfer the SVHDXflt driver to Windows 11, which allows you to create a shared VHDS disk between several VMs. But that's a completely different story.