Citrix Provisioning (PVS) and Machine Creation Services (MCS)
MCS: clone from a snapshot
Machine Creation Services asks the hypervisor to clone a master image snapshot, then personalizes each clone with an identity disk carrying the AD password and computer name, plus (for pooled machines) a difference disk that absorbs writes and is discarded at reset. Requirements: a hosting connection with credentials, a master image with the VDA installed, a snapshot, and free storage. On supported storage, fast clone avoids full copies. Updates flow through catalog update: snapshot the new image, apply, choose how many machines at a time, and roll back if needed.
PVS: stream from a vDisk
Provisioning Services boots target devices from a shared vDisk over the network. Components: PVS server (streaming and SOAP services), farm, site, store, vDisk pool, device collection and target device.
- Boot: PXE using DHCP options 66/67, or BDM (Boot Device Manager) to remove the PXE/DHCP dependency-usually preferred in modern networks.
- Standard image mode: many devices share one read-only vDisk; writes go to a write cache (RAM with overflow to disk, or disk).
- Private image mode: one device, persistent writes, no write cache-used for image maintenance, never for a shared collection.
- Versioning: place the vDisk in maintenance mode, boot one maintenance device, update, then promote a new version.
Worked example. 300 pooled desktops on vSphere with no PVS servers and no PXE: choose MCS. It needs only the hosting connection and a snapshot, supports fast clone for speed, and catalog update handles rollout with rollback. Choose PVS instead when you must stream to physical devices, boot bare metal, or serve one image to many collections.
Traps
- Sharing a private-mode vDisk across a collection-writes collide.
- Editing the base vDisk while devices are booted from it instead of promoting a version.
- Undersized RAM write cache causing overflow and stalls.
- Forgetting to power off and snapshot the MCS master before an image update.