For a long time we have known that corporate use of PCs is inefficient and overly expensive: analysts estimate that a typical PC costs around three times as much as its purchase price to manage over the PC's lifetime. But, until recently, there has been little that organizations could do to change the situation while still delivering acceptable service. Virtualization has changed this in a number of important ways: the physical PC need no longer be the key delivery mechanism and, hence, images can be hosted pretty much anywhere. Essentially, we can host copies of a client operating system and deliver a display protocol to users over the network. However, as with many things, the devil and the opportunity are in the details. Let's look in more detail at how cloud-hosted clients can work today and how changes underway will improve the situation in the future.

Today, we can move desktops into the cloud and manage them in much the same way that we currently manage physical machines. A service provider can build a system to deliver client desktops hosted in the cloud. A customer organization can provide copies of its gold build desktops to the provider, who would replicate them for each users and then allocate an image to every user the first time they connect. From that time on the users will be linked to that same image each time they connect. Behind the scenes, the provider will take care of all of the housekeeping, such as:

• Storing the virtual machine for the user
  
• Delivering users' VMs to a hypervisor running on one of the service provider's servers, and starting, stopping and storing the virtual machine
• Dealing with issues of authentication and integration with customer systems
• Managing security of data and communications with the users' virtual machines

The attraction of this system is that it is very similar to the current ways of operating and, therefore, familiar to IT organizations: users own a desktop image, the image is patched with existing tools and, at the end of the virtual machine's life, standard processes can be used for the destruction of the image. Not that implementing such a system is easy. As with any new way of working, there are problems to be solved: in this case, who is going to be responsible for each step of the image lifecycle from creation through patching, maintenance and support to eventual destruction and, ultimately, who takes responsibility for any failures.

The strength of this approach - its similarities to current practices - is also its real weakness: it does not change the model sufficiently enough to really change the economics of client computing. If we look at the details of a solution such as this, on the positive side we see that the customer can benefit from the economies of scale that a cloud provider can bring in terms of offering servers cheaply and, if users are widely dispersed, then the provider's networking strength could deliver a better interactive experience than could be achieved if the customer organization was hosting the servers internally. On the downside, the PC still needs to be managed in much the same way as before. While we have reduced the need for desk-side support, we have added a new layer of administration between the customer organization and the provider. Hence, this solution will play well in situations where there is some additional benefit of moving user desktops to a provider and out of the customer organization, such as numbers of widely dispersed users outside the LAN, but not more generally.