Every PC does not need its own full copy of the same operating system. This tour publishes a template version, creates two PCs from it, writes different data on each and compares the size each PC sees with the space actually used.
What the PC sees as its disk. Disks are thin: space is taken as data is written.
Everything a PC writes after the base image — documents, but also updates, caches and installed software. It grows with use.
Older recovery points keep the data that has since changed. Their space is measured and counts against the owner’s quota.
Distributed storage keeps data as several replicas or with erasure coding (Replicas / erasure coding: profiles 2+1, 2+2, 4+2, 8+3), so raw disk needed is larger than the data.
| Product version | pre-release builds of 2026-10-23 and 2026-10-30, shipped in release 81fce3de6 |
|---|---|
| Image model | template version = protected block-device snapshot on distributed storage; each PC = copy-on-write clone with its own writable disk |
| Result, 2026-10-30 | Windows “save as new template version” passed end to end (BIOS and UEFI); PCs created from the new version completed first boot |
| Result, 2026-10-23 | two automatic snapshots of one PC disk were measured and written back to the owner’s quota |
| Quota policy | per-owner storage quota: warning, then a 24-hour grace period, then the disk is stopped; adding quota or deleting data restores it |
| Pool growth | by adding disks or hosts to the distributed storage cluster |
| Scale | a few PCs; no large-scale test — no savings percentage or PC count is claimed |
Read more: Office PCs and centralized storage · Distributed storage
The sandbox is invite-only and not open yet; there is no public live demo. Please do not send passwords, keys or customer data with a request.