IT managers think of most technologies in the data center as layers. There are the seven layers of the Open Systems...
Interconnection networking model, the layers of server management and the layers of security.
But if you mention layers and desktop virtualization in the same sentence, all sorts of questions arise such as how can you layer a desktop experience, what makes up the layers of a desktop, how can layers improve virtual desktop infrastructure (VDI), and how will layers affect the end-user experience? The best approach to understanding the concept of a layered virtual desktop is to think of it in terms of a layered cake that consists of the elements of a virtual desktop.
The layer-cake model
Simply put, the layer-cake model is the device + operating system + applications + user data. Ideally, each layer (except for the device) is virtualized. This provides deployment flexibility and enhanced agility, as well as a cost savings, since the layers can be replicated and managed independently.
Furthermore, the layer cake allows VDI technologies to provision a virtual machine (VM) and then to assemble the various layers to create the virtual desktop requested by the end user. By having each element virtualized, IT administrators can quickly change operating systems, control which applications are delivered to the desktop and have granular control over the data presented on the virtual desktop.
The layer-cake model is not new. Companies like Citrix have promoted the benefits of such a method for several years through its application streaming products (XenApp) and its virtual desktop products (XenDesktop).
However, what is new is the application of the layer-cake model to VDI.
The differences between a layered VDI management scheme and the layer-cake model used with something like Citrix XenDesktop are subtle, as summed up in the adage, "VDI is desktop virtualization, but desktop virtualization is not VDI." In other words, there are several delivery models for virtual desktops; VDI is just one of them.
Looking at the layer-cake model from that point of view, there are clearly different use cases, each affecting the viability of the model.
The layer-cake approach makes the most sense in a virtualized environment that relies on a persistent connection and uses high-speed links. In this situation, the layers of computing can be controlled in two ways: The VM can be provisioned in real time -- upon user access -- or it can be pre-provisioned prior to end-user availability.
The first method assembles the VM -- OS, apps and user data -- when the user attempts to access a virtual desktop. Although this method causes a delay while the VM is being assembled, the user is guaranteed to have the latest operating system, apps, policies and settings whenever he starts a new session.
In the second method, the layers are used to build virtual machines for later deployment. Users get instant access to their virtual desktops while the administrator has granular control over the OSes deployed and the applications, policies and user data made available. In addition, the admin can selectively apply patches, upgrades and other changes for all available VMs. However, most VDI technologies don't support a layer-cake model. Therefore, to incorporate the application virtualization component into the layer cake, IT needs to look into third-party products, such as AppSense or App-V, which increase costs.
The layer-cake model will be adopted once market forces prove that it reduces the operational costs and management chores associated with VDI and hosted desktops. Until then, the best bet for those looking to layered virtualization to solve their support and expense woes is to wait until the technology matures and becomes a tightly integrated component of a VDI product.
ABOUT THE AUTHOR:
Frank Ohlhorst is an IT journalist who has also served as a network administrator and applications programmer before forming his own computer consulting firm.