A discount VPS page says 80 GB of storage. You install the operating system, upload a site, run df -h, and read 74 GB with 6 GB already consumed by things you cannot see. The advertised number and the delivered number are almost never the same, and the difference is usually larger on cheap plans than on premium ones, because the same fixed overhead is spread across a smaller volume.
Buying storage you cannot use is a common budget hosting mistake. This piece breaks down where the bytes go, what counts as fair overhead versus creative accounting, and how to audit a plan in the first ten minutes after provisioning so the gap is documented before it matters.
Where the missing gigabytes go
Every host reserves space outside your allocation. A filesystem journal, an ext4 or XFS metadata reserve of 5 percent, a boot partition, swap on disk, snapshot deltas and the hypervisor’s copy-on-write layer all consume physical storage that shows up in the advertised figure or does not.
On a typical 40 GB plan, a clean Debian install leaves about 37 GB usable: 4 to 6 percent lost to the journal, 512 MB to 1 GB to boot and swap files, and the rest to filesystem overhead. That is normal. Losing 20 percent is a different conversation, and it usually points to thin provisioning or snapshot retention rather than routine formatting.
Advertised GB versus delivered GB
Providers quote disk in one of three ways, and they are not interchangeable. Raw provisioned size is the block device total. Filesystem size is what df reports after formatting. Usable size is what remains after the OS and your base packages. Only the third number matters for planning.
Hosts that quote raw size and subtract nothing are not lying, but the plan reads better than it performs. A 100 GB raw disk on a copy-on-write storage backend can show 88 GB inside the guest once the reserve is applied. Ask for the guest-visible figure before purchase.
The copy-on-write and snapshot tax
Discount hosts often run shared storage with thin provisioning. Thin provisioning lets them sell more space than exists on the array, which is fine until everyone fills up. The visible effect is that your volume may be smaller than the plan label, or that writes slow sharply once usage crosses a threshold.
Snapshot space is the other hidden tax. Some providers keep hourly snapshots on the same volume, so each snapshot occupies space until it rotates out. A 40 GB plan with six retained snapshots can genuinely hold only 34 GB of new data. Check whether snapshot retention counts against your quota, and whether you can delete individual snapshots to reclaim space without opening a ticket.
What fair overhead looks like
A reasonable benchmark for a 40 to 80 GB plan is 92 to 96 percent usable after a minimal Linux install. Below 90 percent, ask why. On larger volumes of 200 GB or more, 97 percent usable is common because fixed overhead is a smaller share of the total.
If a host advertises 50 GB, delivers 45 GB, and documents the journal and reserve on the plan page, that is honest engineering. If the same gap exists with no explanation, treat it as a sign that other specs may also be optimistic.
Auditing a fresh install in ten minutes
After provisioning, run lsblk to see the block device size, df -h to see the mounted filesystem, and du -sh /usr /var /opt to find heavy default packages. Compare the block device figure to the advertised one first, then the mounted figure, then free space after your install.
Record the three numbers in a note with the date. If you later raise a ticket, a clear before-and-after comparison resolves the discussion quickly. It also gives you a real figure to put next to competitor plans in your own comparison, instead of the marketing label from every pricing page.
Choosing a plan with the right headroom
Plan storage as usable space, not advertised capacity. If your site needs 30 GB in twelve months, buy a 60 GB plan on a 90 percent-usable host or an 80 GB plan on a 60 percent-usable host. Apply the same rule you would for RAM headroom and the arithmetic gets simple.
Also check the upgrade path. Hosts that resize without a rebuild let you add disk later; hosts that require migration make under-buying expensive. Our VPS plan breakdown notes which plans allow in-place resize, which matters more than a few advertised gigabytes.
When a disk quota gap is a deal-breaker
Three situations make the gap unacceptable. First, a provider that advertises a size the guest cannot ever reach. Second, quota enforcement that locks writes at 90 percent instead of 100 percent. Third, storage that throttles to SATA speeds after a soft limit, changing the plan’s performance without changing its price.
Outside those cases, a 5 to 8 percent gap is ordinary and not worth switching hosts over. The goal is not to find a host with zero overhead, because none exists. It is to know the real number so your capacity planning, backup sizing and upgrade timing are all based on a figure that exists rather than a figure on a sales page.
Before you renew, check the delivered figures on your current plan, then compare them honestly in our budget VPS comparison. A discount plan with a documented 93 percent usable disk is a better buy than a full-price plan with an unexplained 15 percent shortfall.
