The logic of cloud-hosted desktops
Those of you who track Desktone and this blog know that lately we've been spending more time talking about cloud-hosted desktops. This is obviously due in part to the incredible dialogue that's surrounded cloud computing over the past year. However, it's also because we believe that cloud-hosted desktops can and will fundamentally change enterprise IT's willingness and ability to leverage cloud computing in the near term.
The short version of our theory goes like this: Although most discussions about enterprise cloud computing to date have centered around virtualizing server workloads and moving them to a service provider cloud, the challenge of dealing with the data associated with those servers appears to be inhibiting the adoption of this model. That is, in the majority of cases, if you want to move a server to the cloud you have to move the related data with it. This is a non-starter for many enterprises concerned about data security in a cloud infrastructure.
Desktops, on the other hand, have become increasingly separated from their data. More and more enterprise data (i.e. security credentials, mailboxes, home directories, etc.) lives on networked storage. Therefore, enterprises can virtualize desktops and move them to the cloud without moving the data processed on those systems. This is a critical factor in the decision process and one we believe will help accelerate the adoption of cloud computing in the enterprise. In fact, we believe that the desktop will not only lead the charge but will also drive adoption of virtualized servers to the cloud.
For the detailed version of the story, please see an article that I just wrote for SYS-CON's Cloud Computing Journal.
We've only recently started publically pitching this concept, with the first real showcase being our webinar on cloud-hosted desktops. However, support is already starting to show up in both the analyst and press communities. Here's what Rachel Chalmer's from The 451 had to say about our theory:
What Desktone proposes is that, while organizations would like to push servers into partner clouds, in practice server workloads are too tightly bound to sensitive corporate data to make that an attractive proposition. In desktop workloads, data is already decoupled and hosted elsewhere: authentication and authorization in Active Directory, for example, and business data in applications on a server. Desktone argues that this makes it easier to host desktops in the cloud than to host servers there. IT shops can keep the sensitive data where it is, while hosters take on the PCs. It's a strong story...
SoftBank Telecom offers desktops as a service with Desktone's Virtual-D platform
10/29/08
(Note: Access to this report requires a 451 subscription)
Similarly, in a recent article, David Webb of Network World Canada starts off by saying that people are concerned with the security risks of the cloud. However, he then quotes security expert Craig Balding, who says that there are also security benefits to the cloud:
Most significant is the centralization of data, he says. “Large corporations have a problem with asset protection,” Balding says. That’s because there can be instances of the data in a number of places, including employee laptops.
“People with thick clients are bound to download files,” he says. And all too often, those thick-client laptops hold unencrypted data.
“If you combine cloud computing with thin clients,” which only hold small amounts of data in cache, there is less physical exposure to data leakage.
“I think the cloud providers have got an argument there,” Balding says.
Interesting that cloud-hosted desktops are the first security benefit cited in the article, which goes on to talk about other ones such as improved logging and incident response capabilities. I'd be interested in knowing your opinion on this subject and whether you agree or disagree with our theory.

Reader Comments (1)
Jeff, interesting theory. And I have to agree.
However, since you mentioned "networked storage" (Vs traditional SCSI/FC/SAS server<->data relationship) we might even foresee technologies and practices that allows you to decouple (efficiently?) the back-end service running inside a virtual server that in turns run on top of a physical system... where the data are logically connected to that back-end service/virtual server but are in practice located at the customer premises.
After all virtualization is all about segmenting physical servers / networking / storage (i.e. OS/application code + data).....
See the first part of this post: http://it20.info/blogs/main/archive/2008/11/14/162.aspx
By chance the BladeCenter S discusses a use case that is exactly the opposite of what I am trying to argue here. The BC S is all about integrating all these components/subsystems into a single box. Perhaps there is a way for these components/subsyetms to work in a "dispersed scenario" for other use cases (i.e. cloud computing).
Example: the cloud provider could host physical servers whereas OS/application code + data remains hosted at the customer's site. Or alternatively the cloud provider could host physical servers + OS/application code whereas the data portion remains hosted at the customer's site.
In the first scenario the cloud provider would only provide CPU/Memory cycles.... in the second scenario the cloud provider would provide CPU/Memory cycle + storage to host the VM (w/o the data).
Hard to think about something like this in a traditional world but perhaps 10Gbit Ethernet and this "converged network" concept for which Ethernet is to become the highway for IP and Storage traffic ... it will open up new possibilities.
Massimo.