About this article: This article is reviewed against the cited sources. This is source research, not lab validation. Check the stated versions and your own requirements before making changes.
A thick disk reserves all its space on the shared storage when it is created. A thin disk reserves space only as data is written, so the provider can promise more disk space in total than physically exists. Either way, every server's disks sit in a shared storage pool, and if that pool fills, a server writing to it can stop responding even though Windows still shows free space. Thick disks do not make the pool immune: snapshots and other servers' growth use the same space. When comparing hosting quotes, ask two things: are you billed on the size you were given or the space you use, and who watches the shared storage so it never fills?
Why it matters
Storage can be a large line on a hosting quote, and "500 GB" can mean different things. On a thick disk, 500 GB is set aside for you. On a thin disk, 500 GB is a ceiling, and the platform only holds what you have actually written. That difference can affect price, and it affects risk: an estate of thin disks depends on someone keeping the shared storage below full.
How it works
The three VMware disk formats
Broadcom's overview of the disk formats supported by ESXi describes them:
- Thick Provision Lazy Zeroed (the default): "Space required for the virtual disk is allocated during the creation of the disk file." Old data on the physical storage is cleared when first written.
- Thick Provision Eager Zeroed: space is allocated at creation and cleared immediately, and "It might take much longer to create disks in this format than to create other types of disks."
- Thin Provision: "Space required for the virtual disk is not allocated during creation, but is supplied and zeroed out, on demand at a later time."
What happens when shared storage fills
The datastore is the shared storage pool that holds many servers' disks. Broadcom's article on this situation, written for ESXi 7.x and 8.x, states that "A virtual machine stops responding when a datastore exhausts its physical capacity." It adds that this "often occurs during thin-provisioned disk growth or snapshot consolidation", and that "the Guest OS reports free space while the vSphere datastore indicates it is full".
That last point is why this is the provider's job to watch: your server cannot see the problem coming. Its own disk looks half empty while the pool beneath it is full.
A thick disk's own data area is reserved, so the disk itself does not need to grow. But the cause Broadcom names is the pool running out, and snapshot files and other servers' thin disks draw on the same pool. Thick disks lower one source of growth; they do not remove the need to watch the pool. A thick format is also a reservation at the virtual disk and datastore layer only: if the storage array underneath is itself thin provisioned, the physical space may still not be set aside, so ask how the array is provisioned too.
Deleted files and reclaimed space
When you delete files in Windows on a thin disk, whether the space goes back to the shared pool depends on every layer between Windows and the physical storage supporting reclamation, and on the provider's settings. We could not find a public source showing how a particular hosting platform handles this, so ask your provider whether deleting data shrinks what the platform holds for you, and whether your bill follows.
What to choose and why
The table is our editorial summary of the Broadcom sources above; billing depends entirely on the provider.
| Thick | Thin | |
|---|---|---|
| Space reserved on shared storage | All of it, at creation | Only what is written |
| Effect of a full pool | The disk itself need not grow, but snapshots and the server still depend on the pool | Disk growth fails and the server can stop |
| Efficient use of storage | Lower | Higher |
| Billing (ask) | Possibly the full size | Possibly the full size or the space used |
Our view: thin disks are a reasonable choice for a business server when the provider monitors the shared pool and keeps headroom. Thick disks make sense where the workload must never wait on storage allocation, or where the vendor or a clustering feature requires it.
Questions to ask your provider
- Are my disks thin or thick?
- Am I billed on the size allocated to me or the space I use?
- Is the shared storage overcommitted, and by how much at most?
- Who is alerted, and at what level, before the shared storage fills?
- Does deleting data reduce what the platform holds for me, and what I pay?
A hypothetical example
An architecture practice is quoted "2 TB" of storage. It writes 900 GB in the first year. If the disk is thick and billed on allocation, it pays for 2 TB from day one. If thin and billed on use, it pays for 900 GB, but depends on the provider keeping the pool from filling. Either arrangement can work; what matters is knowing which one you bought. The example is illustrative.
Read next
Related articles you may want next.
- HostingDeleting Guest Files Did Not Shrink the Datastore: Trace Discard SupportMap where reclaim requests can travel and determine which layer owns retained allocation.Read article
- HostingGrowing a hosted disk: extend easily, shrink rarely, and plan the volume layoutWhy a hosted server's disk can usually be made bigger on request but not smaller, and how to lay out disks and volumes so future growth stays a simple, daytime change.Read article
- TroubleshootingWindows Disk Full, VMware Space Free: Map the LayersMap Windows volumes to virtual disks and datastores, then identify where capacity is missing before approving an expansion or filesystem cleanup.Read article
Risks and limits
- A full pool stops servers. Thin provisioning shifts more of the risk to the provider's monitoring; ask how it is done.
- Windows cannot warn you. The server's own free space says nothing about the pool beneath it.
- Snapshots consume pool space whatever the disk format.
- Reclamation is not guaranteed. Deleting files may not return space or reduce charges.
- This article covers VMware ESXi formats. Other platforms use different names and behaviour.
Sources
Opened 15 November 2026; reopened 16 November 2026.
- Broadcom Knowledge: Types of supported Virtual Disks on ESXi hosts, article 308992 (page metadata: modified 9 January 2026).
- Broadcom Knowledge: Virtual machine stops with "There is no more space for virtual disk" on NFS or VMFS Datastores, article 318352 (page metadata: modified 12 November 2026).
Related article: Windows Disk Full, VMware Space Free: Map the Layers covers the opposite case, where Windows is full but the platform has room.
Related service: Cloud hosting.
Learning notes
- Thick disks reserve all their space up front; thin disks reserve it as data is written.
- A full shared pool can stop a server; thick disks reduce one cause but are not immune.
- Windows can show free space while the pool beneath is full.
- Deleting files does not always return space or reduce cost; ask.
- Ask whether billing is on allocated or used space, and who watches the pool.
This article was originally published on .