Every technology challenge appears to have several solutions that often seem unrelated or even contradictory. IT planners for enterprises and small and medium-sized businesses can't chase trends in all directions. It's important to understand how strategies will eventually converge and harmonize to establish a preferred direction of evolution for applications and productivity-enhancement projects and to ensure that IT budgets are used optimally.
Cost management has led to two technical developments: the centralization of desktop computing through
Since cloud computing is a data center architecture, and virtual desktops are hosted in cloud data centers, the combination of these trends is leading the market to desktop as a service or DaaS.
But what exactly does this mean?
Cloud computing creates and manages a server-side resource pool -- the cloud -- while VDI creates and manages a client-side resource pool -- the virtual desktop. The basic technology for these services is the same. Server virtualization creates virtual systems that support applications (cloud computing) or users (virtual desktops).
But the main difference between the two is that cloud computing targets applications designed to run from a centralized environment, while VDI targets those designed to run on the user's local system.
Put simply, VDI is a form of cloud computing where the client/server protocol used is the virtual device interface that mirrors the graphical user interface. Therefore, VDI can be considered a DaaS application of the cloud, and in time, VDI and cloud computing will converge further.
The difficulties of DaaS
VDI and cloud computing start with a pool of hardware resources that can be managed efficiently and shared among applications and users. The two services assume that users have local devices that provide them application access and that these devices are linked to the central pool of hardware resources by the network.
The primary difference between the two in terms of technical requirements is the level of per-user customization.
Applications for the cloud are standardized and run by all users in the same way. There may be per-user customization, but this is based on application parameters and setup -- not on hardware configuration or software requirements at the client or server side.
On the other hand, applications on desktop systems can be personalized to the user. For example, while any set of users can be supported by the same enterprise resource planning (ERP) system, there may be hundreds of configurations of virtual desktops to reflect the mixture of software tools running on users' PCs.
DaaS must, in the limiting case, let every desktop appear as a separate application. Managing a larger number of machine images is the first priority of DaaS; the second is managing the policies to assign resources.
With cloud server applications, the virtual machine image is normally linked 1-to-1 to an application or application component. The resource requirement of that application is largely static, so its effect on other applications is predictable.
With DaaS, the machine image is linked to user applications that may have widely varying resource needs. It's important to properly assess these demands because they could radically affect the performance of other apps whose virtual machines share the same physical resources.
Another potential complication to consider is that a VDI application may be a client for a cloud-computing application.
If the resource assignment for the virtual desktop doesn't consider that the application might be using another cloud application as a service, the virtual desktop and app could be hosted in locations either too far apart to be efficiently connected. Or, they could be on the same server and thus competing with each other.
Power users are the most likely to create these kinds of complications in DaaS. You may want to consider a dedicated resource pools for these users' virtual desktops, or you may want to allow the most demanding power users to retain traditional desktop systems.
Remember that all virtualization technologies assume that system resources are not fully utilized -- that's why multiple virtual systems can share a single physical one. Applications that are likely to fully load system resources are poor candidates for virtual hosting or DaaS.
If we assumed that all applications began to shift to a cloud implementation using SaaS and SOA principles, the "desktop" would increasingly become a thin client. As such, it would have a smaller number of variable software requirements and thus lower support costs. More of what today is VDI would become part of the cloud server application set, making the issues of DaaS management and hosting less different from management of cloud resources. The question is, will that make DaaS widespread by making it easy, or stamp it out because thin-client support is trivial?
In the near future, the multiplicity of thin-client technologies seems to argue for more DaaS. Adobe and Microsoft offers it own thin-client architectures, and various standards for thin-client support are developing. In addition, many workers need to use computers while untethered from the network. As result, they can't use purely hosted versions of software. Hosted versions of productivity software still fall far short in terms of features, so companies moving toward cloud computing should consider DaaS issues as part of their planning activity.
ABOUT THE AUTHOR:
Tom Nolle is president of CIMI Corp., a strategic consulting firm specializing in telecommunications and data communications since 1982. Nolle is a member of the IEEE, ACM, Telemanagement Forum and the IPsphere Forum, and he is the publisher of "Netwatcher," a journal in advanced telecommunications strategy issues.
This was first published in April 2010