Gabe Knuth predicted 2014 as the year of DaaS (Desktops as a Service). Last week he wrote an article confirming that he believed that 2014 was indeed the year of DaaS. Sure VMware acquired Desktone, Amazon released a product called Amazon Workspaces. But does releasing a product make DaaS any more real?
VDI is a niche solution because it (a) has a high total cost of ownership (b) introduces a poor user experience. We believe that DaaS is an even more niche solution than VDI.
Lets first look at the total cost of ownership of a DaaS desktop. An Amazon Performance desktop costs $60/desktop/month. A comparable physical desktop costs $500. One time. Plus the Amazon desktop comes without a monitor, keyboard, or mouse. You still have to buy that. Ok, so a DaaS desktop is not cheaper than a physical desktop. (Neither price includes the cost of managing the desktops - installing patches, updates, applications, etc. - which according to all estimates is still the biggest chunk of TCO.)
What about the user experience when the desktop runs in the cloud? There is latency (x milliseconds) between the typical client and server in a client-server architecture. See figure below.
As shown in the figure below, a virtual desktop even in your own data center introduces an extra hop and extra latency. The client part of the client-server now runs in the data center and the user interacts with a new "client". The overall latency (a+b) is greater than x.
This greater latency significantly affects end user experience for all applications. The fundamental problem is that the laws of Physics for latency cannot be changed. The two special kinds of applications that are most affected by latency and the extra hop are:In most cases, for real-time and multi-media applications the only solution is to route that traffic so that the traffic does not go through the extra hop. This means that the end point needs to have more processing capabilities and requires end point maintenance, thus rendering moot some of the advantages of VDI. Plus this kind of processing requires in most cases that the end point be a Windows end point, which re-introduces all the maintenance and costs of a Windows end point.
If that virtual desktop runs in a cloud, that latency and extra hop is likely even more problematic. Since you are running the desktop in somebody else's cloud, it is likely that (a+b) is even larger. This all but guarantees that the problems will only be worse.
Running your desktops in somebody else's network introduces an additional challenge. How does that desktop get access to your internal network? The server components of your client server application will still run in your data center, so how does the client running in the cloud desktop get access to the server?
So, if you get even worse user experience than VDI at a total cost of ownership that is higher than your physical desktops, why would DaaS be adopted? There are use cases where high TCO and poor user experience is acceptable because there is no alternate solution. For example, if you want to deliver a corporate desktop to contractors in a different country.
If VDI is a niche solution, DaaS is even more of a niche solution. We believe that the right solution for enterprise mobility is just below.