Guide · Storage
Store the operating system once as a read-only template, and give every PC its own writable disk on top of it. Unchanged data is read from the shared template, and only what a PC changes takes extra space. Nobody shares a writable system disk, so one user's changes never show up on another PC.
Imagine an accounting office with 50 identical Windows PCs. Installing Windows and the office software 50 times stores the same few tens of gigabytes 50 times. A shared image avoids that: you prepare the system once, and the PCs are built from it.
These numbers are an illustration to show the arithmetic, not a measurement and not a purchasing recommendation. Units are decimal (1 TB = 1,000 GB).
Assumptions: 50 PCs, a base image of 40 GB, and 15 GB of changes per PC (user files, updates, caches). The storage keeps three replicas of everything.
| Step | Every PC keeps a full copy | Shared base + clones |
|---|---|---|
| Data stored | 50 × (40 + 15) = 2,750 GB | 40 + 50 × 15 = 790 GB |
| With three replicas | 2,750 × 3 = 8,250 GB | 790 × 3 = 2,370 GB |
| If normal use stays at 75% of raw capacity | 8,250 ÷ 0.75 = 11,000 GB | 2,370 ÷ 0.75 = 3,160 GB |
Not included: snapshot history, metadata, extra template versions, allocation granularity, uneven data placement and the headroom you need to recover from a failed disk. The 75% figure is only an example: the right headroom depends on your failure domains and how quickly you want to recover. Every clone also grows with use, because updates, caches and the page file count as changes.
Four different numbers get confused in storage planning. Keep them apart.

When an owner reaches the quota, Nodekeeper warns first, then allows a 24-hour grace period, and then stops the disk. Adding quota or deleting data brings it back. The pool itself grows by adding disks or hosts to the cluster.
| Replicas | Snapshots | Backups | |
|---|---|---|---|
| Protect against | A failed disk or host | A bad update or a mistake on one PC | Losing the cluster itself |
| Where the copy lives | Inside the cluster | Inside the cluster | Outside it |
| Bad data | Copied as faithfully as good data | Rolled back to the earlier state | Restored from an earlier copy |
| Takes space | Multiplies raw capacity | Changed blocks since the snapshot | Space in the backup destination |
One base image per department or hardware group is easier to look after than one per person. A new template version does not upgrade the PCs that were cloned from an older one, and a version that still has clones depending on it cannot be deleted, so decide how you will move people to a newer version before you publish many of them.
Choose the disk size when you create the PC. A larger allocation does not create physical space: it uses quota, and the cluster still needs real capacity behind it. Windows extends the C: drive to the full disk size on first boot.
Retention drives snapshot space. Keep enough history to cover the mistakes you actually see, and add a backup for the case where the cluster itself is lost.
Shared images run on the same central storage that carries your virtual machines and member files, so one panel covers capacity, quotas and recovery for all of them. Windows and Linux office PCs boot from this storage over the network, as explained in What is Windows network boot?.
Shared images and distributed storage are part of the Business plan, licensed per node (each machine that runs Nodekeeper).
Block, file and S3-compatible object storage, quotas and recovery in one place. Start with the storage overview, or talk to us about sizing your pool.
30-day trial of Business: monthly subscription, first month $0, card required, once per account, up to 3 nodes. Cancel before day 30 and you are not charged; after that, Business renews at $29 per node per month.
Manage websites, office PCs and servers from one panel on your own hardware.