Images, applications and user state
Image design balances compatibility, isolation, support effort, update control and user needs. A single universal image may be attractive operationally but unsuitable for applications requiring different operating systems, drivers or security controls. Conversely, an image for every individual can create unnecessary maintenance. Use an application and workload assessment to decide which requirements can share a supported base and which need a separate delivery design.
Provisioning choices
Machine Creation Services uses hosting-platform integration and a prepared image to provision machines according to the selected catalog and platform. Citrix Provisioning streams a vDisk to target devices and introduces its own streaming, bootstrap, network and write-cache dependencies. These approaches have different operational trade-offs; neither is always superior. Size and validate storage, network, cache and image-update behaviour using representative workloads and the supported implementation. A boot-device method can replace a PXE/TFTP bootstrap path without proving that every use of DHCP has disappeared; target IP configuration is a separate question.
Persistence is a separate requirement
Random or dedicated assignment describes which machine a user receives. It does not alone establish when every disk or user setting resets. Define machine persistence, restart behaviour, image-update behaviour and the location of personal state separately. Profiles, redirected data and application-specific caches may need different solutions. Plan access permissions, availability, size, concurrency, backup and migration. A profile container is not a backup and is not permission to write the same state concurrently from unsupported sessions.
Layer and deliver applications deliberately
App Layering can separate OS, platform and application maintenance and use supported image templates or runtime attachment approaches. Verify compatibility and the particular layer type before claiming that every application can be elastically delivered. Drivers, services, application isolation and platform dependencies can justify separate images or another delivery mechanism. WEM can apply user-environment actions and resource-management features, but changes still require compatibility testing and evidence of their effect.
Control updates and rollback
Maintain a reproducible source image and record changes. Test an update with representative applications and users before staged production rollout. Know what rollback restores and which user data lies outside it. A previous image snapshot cannot automatically restore lost profile data or repair every infrastructure change. For a hypothetical migration, identify who owns the data copy, how it is verified, when the cutover occurs and how the prior state remains recoverable. Re-baseline capacity after meaningful image or application changes.