However,  giving users a whole virtual desktop so they can run a couple of extra applications remotely isn’t always desirable. Sometimes it would be ideal to publish an application that is installed on a virtual desktop to a user, so they can run that application without also presenting them a whole desktop. Remember back to when  Citrix introduced seamless published applications – so much of that complexity went away for the user when they could just run the application, without also having to interact with a second desktop session.

Another usage scenario applies to thin terminal users – the users today that do run a whole published desktop from a multi-user server OS. Administrators could provide those users a published application within their published desktop – but instead of running on the server desktop, the application runs from a workstation virtual desktop, and is presented into the user’s server desktop session. This would effectively allow more applications to be provided in the thin client environment, without switching over to a completely virtual desktop environment. In essence, it would allow you to run some server based pulished apps, and some desktop based published apps – whatever was needed and whatever made sense for the administrators.

So, sounds like something that makes sense, right?  Citrix thought so, and have recently announced this new feature (planned for Q3 2009) that allows you to publish applications that are running on virtual desktops. These applications can then be run remotely, just like any other XenApp seemless published application.

Citrix has called the feature VM Hosted Apps, and it will be a great addition to the XenApp suite which will give orgainisations another option for those applications that don’t like running in a multi-user, or server based environment.

Of course Quest has had this capability for a while now with via their vWorkspace product.  In fact Quest has long been promoting a use case for virtual desktops along these lines – using the single user workstation OS to run applications that the terminal servers can’t – but publish the apps through to the terminal server environment where user density is more cost effective.

And I should also mention Microsoft’s RAIL – Remote Applications Integrated Locally. It’s essentially the same concept also, and is used in Windows 7 to publish applications that are running on the “XP Mode” virtual desktop through to the Windows 7 start menu and desktop.

Whichever way you look at it, publishing applications from a variety of machines and OS’s, and running them in a variety of combinations and locations will soon be possible from multiple vendors. More options. More flexibility. More complexity?  The question is, how will we manage and assign apps? But that’s a conversation for another day.

Centralised Virtual Desktops

Centralised Virtual Desktops (or Virtual Desktops hosted on central Infrastructure – VDI) have a range of benefits today.  For a start, most VDI implementations force you to separate the OS, from the apps, from the user data and settings. That in itself provides greater flexibility and more granular options for quick troubleshooting and repair (each layer can be treated separately by IT). And the fact that the virtual desktops are usually hosted in a central data centre that is “near” the data and application back-ends enables much of the layering to work efficiently, and can provide a better user experience in some cases for remote offices/home workers. And of course it also allows orgainisations to control their data and applications more closely – they never leave their network.

However, the number of VDI implementations in the world are relatively small. When many of my customers have looked to implement VDI solutions, they have found it doesn’t suite the requirements of the whole organisation. Now to be clear, I have nothing against VDI – in fact I think virtual desktops are great! Problem is, many people still have really valid reasons why centralised (or remote) virtual desktops aren’t always what they need.

Distributed Virtual Desktops

And that’s where client side hyper visors come in.  Client hyper visors will allow the virtual desktop to be run on user hardware (like a laptop), rather than on centralised servers. So it can be run offline, and you don’t have the constraints that remote video/audio imply.

So what is good about the client OS running virtually? Essentially, it’s about abstraction of the OS from the hardware. Once we have decoupled the OS, it becomes really easy to take snapshots, incremental backups, restore those snapshots/backups, or move the computer to new hardware. And speaking of moving the OS to new hardware – we should expect that most of the client hyper visors will have the ability to sync changes made locally up to a central copy, so that you can actually run that same environment from a variety of devices.  When you log off your computer at work, all changes are sync’d up to the network, so you can use the same virtual machine remotely from home  – this time running from a server hyper visor in your data centre – or in someone else’s data center on the internet (in the cloud). And changes made from home will be sync’d back down to your work computer when you turn it on next. Or if you loose your laptop and get a new one, then the whole virtual machine will be downloaded and you will be right where you left off.

The other great advantage of the client OS running virtually is that you can run multiple client OS’ at the same time – without any one of them being the ‘master’ or host. A typical example the vendors are talking about is running a personal desktop and work desktop. But there are many other scenarios, like different SOE versions, test machines, developer machines, etc.

Management

Both Citrix and VMware recognise the need for the client hyper visors to have management capabilities – these will be crucial to the successful adoption of client side desktop virtualisation. IT departments will need to be able to manage OS assignment, push, and sync. And controlling which devices/storage/networks the various virtual machines have access to will be critical in controlling the security and reliability of the environment.

After the market matures a little, the exact type of hyper-visor implemented will not be the major factor when choosing a client hyper-visor vendor. Instead, the  management capability, third party extension/plugin model, and integration with existing systems and processes (e.g. Microsoft Configuration Manager and Active Directory) will be the key differentiators orgainisations will be considering when choosing a product set.