Thin vs Thick Provisioning: Complete FAQ Guide

VMware admins ask the same questions about thin and thick provisioning repeatedly—which one’s faster, can you convert between them, what happens when storage fills up. This FAQ answers those questions with practical guidance based on real-world VMware deployments.

What you’ll find here:
Direct answers to the most common provisioning questions, plus expanded explanations that help you make better storage decisions for your specific environment.

Quick Answer Summary

What’s the main difference?
Thin provisioning allocates storage on-demand as data is written, while thick provisioning reserves the full disk capacity upfront when you create the VM.

Which one is faster?
Thick eager zeroed provides the most consistent performance, especially for production databases on traditional storage arrays.

Can I switch between them?
Yes, you can convert between thin and thick provisioning using Storage vMotion without VM downtime.

What Is Thin Provisioning in VMware?

Thin provisioning creates a virtual disk that grows as your VM writes data to it. When you provision a 100GB thin disk, VMware creates a small VMDK file that expands only when the guest OS actually stores data.

This approach maximizes storage efficiency because you can allocate more virtual disk space than you physically have available. A 10TB datastore can support 20TB of thin-provisioned virtual disks if the VMs only use 8TB of actual data. You’re essentially over-provisioning your storage capacity.

The trade-off is monitoring complexity. You need to track both provisioned capacity (what you’ve allocated) and used capacity (what’s actually consumed). Without proper monitoring, your datastore can fill unexpectedly when multiple VMs grow simultaneously.

What Is Thick Provisioning in VMware?

Thick provisioning pre-allocates the entire disk space when you create the virtual machine. A 100GB thick disk immediately consumes 100GB on your datastore, regardless of how much data the VM actually stores.

VMware offers two thick provisioning types. Lazy zeroed thick reserves space immediately but only writes zeros when the VM writes data to each block. Eager zeroed thick reserves space AND writes zeros to every block upfront, which takes longer during VM creation but delivers the best performance.

Thick provisioning eliminates over-provisioning risk and simplifies capacity planning. What you see in datastore usage is exactly what your VMs have reserved. You can’t accidentally over-commit storage capacity.

Which Provisioning Method Offers Better Performance?

Thick eager zeroed delivers the most consistent write performance because blocks are already zeroed. There’s no penalty on first write to each block, which matters for high-IOPS workloads like production databases.

Thin provisioning performance depends heavily on your storage array architecture. Modern all-flash arrays handle thin provisioning with minimal overhead—the performance difference compared to thick provisioning is often negligible. Traditional spinning disk arrays show more measurable performance degradation with thin provisioning due to metadata updates and potential fragmentation.

For production SQL Server, Oracle, or PostgreSQL databases on traditional SAN storage, use thick eager zeroed. For most other workloads on modern storage platforms, the performance difference won’t justify the reduced storage efficiency of thick provisioning.

Can I Convert Between Thin and Thick Provisioning?

Yes, VMware lets you convert between provisioning types using Storage vMotion. You can migrate a thin disk to thick (or vice versa) without VM downtime.

The conversion happens during the storage migration process. When you Storage vMotion a VM, you can select the destination disk format—thin, thick lazy zeroed, or thick eager zeroed. VMware handles the conversion automatically as it copies the VMDK to the new location.

Keep in mind that converting from thin to thick eager zeroed takes longer because VMware must write zeros to all allocated blocks. Converting from thick to thin is faster but doesn’t immediately reclaim space—the thin disk starts at the same size as the thick disk and only shrinks if you enable UNMAP operations.

What Happens When a Thin-Provisioned Datastore Fills Up?

When a thin-provisioned datastore reaches 100% capacity, VMs can’t write new data. You’ll see write errors in guest operating systems, applications may crash, and you can’t create new VMs or snapshots on that datastore.

This is why monitoring is critical for thin provisioning. Set alerts at 80% datastore capacity so you have time to add storage, migrate VMs, or delete unnecessary snapshots before you hit the limit.

Tools like PRTG’s VMware datastore sensors track both provisioned and used capacity, alerting you before capacity issues cause outages. You’ll see trends showing which VMs are growing fastest and when you’ll likely run out of space.

When Should I Use Thin Provisioning?

Use thin provisioning when storage efficiency matters more than absolute maximum performance. Development and test environments are ideal candidates—you can spin up VMs quickly without committing storage capacity upfront.

VDI deployments benefit significantly from thin provisioning. Desktop VMs rarely use their full allocated disk space, so you can over-provision desktop pools and serve more users with the same physical storage.

File servers and non-critical workloads also work well with thin provisioning, especially on modern all-flash arrays where performance differences are minimal. Just make sure you have robust monitoring in place to track datastore capacity and growth trends.

When Should I Use Thick Provisioning?

Use thick provisioning for production databases and mission-critical applications where consistent performance is non-negotiable. SQL Server, Oracle, and PostgreSQL workloads benefit from thick eager zeroed provisioning, especially on traditional SAN storage.

VMware Fault Tolerance requires thick eager zeroed disks—you can’t enable FT on thin or lazy zeroed virtual disks. If you’re planning to use FT for high-availability VMs, you must provision them as thick eager zeroed.

Environments without robust monitoring tools should also favor thick provisioning. If you can’t track provisioned vs. used capacity effectively, thick provisioning prevents the risk of unexpected datastore exhaustion.

Does Thin Provisioning Work With VMware Snapshots?

Yes, thin provisioning works with VMware snapshots, but snapshots behave differently than on thick-provisioned disks. When you create a snapshot on a thin-provisioned VM, the delta disk is also thin-provisioned.

The challenge is that snapshots can cause thin disks to grow rapidly. Every write to the VM goes to the delta disk, which expands the overall storage consumption. If you leave snapshots in place for extended periods, your thin-provisioned VMs can fill the datastore faster than expected.

Best practice: Delete snapshots within 24-72 hours. Don’t use snapshots as long-term backups. Monitor datastore capacity closely when you have active snapshots on thin-provisioned VMs.

How Much Can I Over-Provision With Thin Provisioning?

There’s no hard technical limit, but best practice suggests staying under a 2:1 over-provisioning ratio. If you have 10TB of physical storage, don’t provision more than 20TB of thin virtual disks.

Your actual safe over-provisioning ratio depends on your workload characteristics. VDI environments with predictable desktop usage patterns can safely over-provision 3:1 or even 4:1. Production server environments with unpredictable growth should stay closer to 1.5:1.

Monitor actual usage trends for at least 30 days before deciding on your over-provisioning ratio. Track how fast VMs grow, identify seasonal patterns, and build in a safety margin. Always set capacity alerts at 80% full to give yourself time to respond.

Still Have Questions?

For more detailed guidance on VMware storage management and monitoring, check out Paessler’s VMware monitoring resources. You’ll find best practices for tracking datastore capacity, performance metrics, and storage efficiency across your virtual infrastructure.

If you’re planning a storage migration or evaluating provisioning strategies for a new VMware deployment, test both thin and thick provisioning with your actual workloads on your specific storage platform. Performance characteristics vary significantly between storage vendors and array models.