95% of performance issues in persistent VDI can be traced back to storage I/O contention or capacity exhaustion.


The Persistence Paradox

Citrix Machine Creation Services (MCS) excels at automating VDI deployment. For Static VDI, where each user has a dedicated machine, MCS provides a persistent experience, saving all user changes. However, this creates a significant challenge: each VM's storage footprint continuously grows, demanding proactive and intelligent capacity management to prevent performance bottlenecks and service outages.


Anatomy of a Static VDI Disk

  • 1
    1. Base Disk: A read-only copy of the master image, shared by all VMs on a single datastore to save space.
  • 2
    2. Difference Disk: The workhorse of persistence. This unique, thin-provisioned disk stores all user changes and grows over time, consuming the most unique space per VM.
  • 3
    3. Identity Disk: A tiny (16MB) but vital disk that gives each VM its unique identity within Active Directory.
  • 4
    4. Write-Back Cache (Optional): A non-persistent disk used by MCS I/O to optimize performance by caching write operations.



VM Provisioning Flow: How MCS Allocates Storage


When you configure a hosting connection with multiple datastores, MCS uses a specific strategy to place new VMs. This flow demonstrates the default "Enhanced Round-Robin" logic.



The Challenge: When a Datastore Fills Up

MCS allocation logic only applies to *new* VMs. It does not automatically rebalance existing VMs. This is what happens when a datastore's capacity is exhausted by growing difference disks.Visualization of datastore utilization

Impact & Resolution

New Provisioning Halts

MCS detects that Datastore A is full and will no longer place new VMs on it. Future VMs are automatically directed to Datastores B and C.

Existing VMs Unaffected (Initially)

VMs already on Datastore A remain operational but may face performance issues or fail if they need to write more data than space allows.

! Administrator Intervention Required !

MCS does not automatically fix this. Admins must act by:

  • Adding more storage to the hosting connection.
  • Using vMotion to migrate existing VMs off the full datastore.
  • Marking the full datastore as "superseded" in Citrix Studio.

Architect's Choice: Delta Disks vs. Full Clones


For static VDI, MCS offers two provisioning methods. The choice involves a crucial trade-off between storage efficiency and management simplicity.

FeatureFast Copy (Delta Disks)Full Copy (Full Clones)
Storage EfficiencyHigh. Uses a shared base disk, minimizing initial storage footprint.Low. Each VM is a complete, independent copy of the master image.
Management ComplexityMedium. Requires careful monitoring of individual difference disk growth.Low. Simplified management with one primary disk per VM.
Provisioning SpeedFast. Only unique disks need to be created.Slower. A full copy of the entire disk must be created for each VM.
Image UpdatesUpdating the catalog affects all VMs that use the base disk.Updates only apply to *new* machines. Existing VMs are unaffected.