The two sides of the OS
There has been much talk recently about how the OS (read Windows) is becoming less relevant in the era of virtualization and Web 2.0. I wanted to take a closer look at this concept from a desktop virtualization perspective and, of course, discuss how this notion impacts the move toward a VDI model.
At the highest level, there are really two sides to an OS. On the bottom or device side, an OS is basically a device manager. On the top or application side, the OS is an application platform. This is an important distinction, especially on the client side, because as I’ve mentioned in an early posting, the PC is really a Swiss army-type device that serves many purposes. I find it easiest to divide those functions into two general categories based on the device/app border. What is even more interesting is to look at how that boundary impacts use cases, such as personal vs. business computing, and I will touch on this a little later.
First , let’s talk about the reality of the OS (Windows) becoming a relic in the near future. On the server side, I think the argument is definitely more credible. Because there is limited diversity in terms of the numbers of server manufacturers, storage controllers, NICs, etc., it is more likely for a hypervisor developer to be able to support that universe of devices. VMware has clearly proven their ability to do this. Things are definitely harder on the application side and Microsoft still shows its dominance in many different types of server workloads including web, mail, database, etc. However, it is possible to envision how certain types of server applications are at risk of having their underlying OS substituted for an open source alternative. We have even seen the concept of server-side applications that can run directly on a hypervisor without an OS in between, such as BEA’s Liquid VM initiative.
On the client side, things are different. Windows dominates the desktop and there are no real alternatives on the horizon for most businesses and other larger organizations. From a technical perspective, there is much greater diversity in both the number of devices and the number applications that need to be supported. On the device side, there are Windows drivers for literally thousands of peripherals. And on the application side, it’s likely that the number of Windows apps in existence is approaching 100,000. Microsoft has been building its ecosystem for more than 30 years and it’s foolish to think that that the lock-in created by all those drivers and apps is going to disappear anytime soon. It’s going to take decades to cycle through all of that investment.
However, when it comes to VDI models, virtualization does have a profound impact on the two sides of the OS from both a technical and a use case perspective. On the technical side, machine virtualization severs the traditional connection between the two sides of the OS. This means that an OS instance no longer has to manage both devices and apps. A locally installed instance can manage the devices while a centrally hosted session delivers the applications. From a use case perspective, this separation allows users to have device independence while receiving their corporate application environment as a network delivered service.
The separation of the two sides of the OS supports new approaches to corporate computing, such an employee-owned PC model where users are incented to purchase their own devices to which IT delivers a “soft PC” or a virtualized session that contains all the users’ business applications. The division also highlights another important aspect of use cases. Beyond standard peripherals, most of the devices that users connect to their machines fall into the personal (vs. business) computing category (i.e. iPods, digital cameras, etc.) Even support for one of the major exceptions, the Smartphone, is becoming less of an issue as these devices shift to OTA (over the air) synching which removes the need for a physical connection to a user’s PC. The point is that device support is primarily a personal activity that IT is supporting in environments where users are issued corporate owned PCs. By migrating to an employee-owned PC model enabled by a VDI delivery system, organizations can get out of the business of supporting user owned peripherals being connected to corporate owned assets. And for IT, that could prove to be one of the more important aspects of the two sides of the OS.
References (2)
-
Response: Bring Your Own PCJeff Fisher from Desktone has posted some thoughts on desktop virtualization. One interesting scenario -
Response: Bring Your Own PCJeff Fisher from Desktone has posted some thoughts on desktop virtualization. One interesting scenario

Reader Comments (3)
Jeff,
This is a great post. Seperating device hosting from application hosting gets to the heart the desktop virtualization decision.
The logical next step is a well defined device acquisition and support strategy. Devices that are self-hosting such as network printers that support a standard print format such as PDF/Postscript or web-cameras with built in web servers alleviate the driver/device hosting challenges currently limiting certain types of desktop virtualization deployments.
I believe this will become the core issue in controlling support costs as customers increase their reliance on desktop virtualization. In other words, devices that are 'friendly' to desktop virtualization are the key enablers of its adoption.
As for looking at the OS in two parts i agree, but when you look at desktops as a service you actually miss the whole point.
In my oppinion the desktop is irellevant it's just a delivery shell and interface to the really important stuff - the applications a desktop in itself is worth nothing.
So it should say applications as a service instead, you could easily compare it to electricity like this:
you draw it from the wall socket on demand, you will need the right ploug ofcourse (endpoint device) and then you dont really need to care or know about the rest.
What kind of energy source produced the power?
(different backend computing power, virtual, physical, cloud, grid etc.)
What mix of sources created it?
(some applications just runs better on certain platforms so get the best for all)
What path took it to get to me?
(different locations would offer different benefits both economically and enviromentally)
If we stick to the traditional view of desktops and it's OS as the container to be running our applications, we will never realize all the benefits that virtualization and beyond could really give us.
Regards
Thomas
Thomas,
Thanks for your comment. I agree that the ultimate nirvana we are all seeking is the ability to access any application from any system at anytime and to have those applications execute either locally or remotely depending on the particular scenario. However, due to the prevelance of Windows applications and their tight coupling with the OS, it's hard to see how we are going to be able to get to a world of pervasive utility computing anytime soon.
With that said, DaaS offers many of the benefits of the desired model (i.e. anytime, anywhere access - with a connection, of course -, stateless computing from the access device perspective, etc.) without having to wait for every application to be converted into its utility form. Sure it doesn't provide the answer to every application delivery challenge, but we believe that it's an excellent option for many while we wait for nirvana to arrive.
Jeff