Wednesday, 21 July 2010

The future of Desktop Virtualization (VDI)?

If you follow the news about VDI you have probably noticed that the market is quickly heating up and booming with new products, vendors and solutions.

The old remote desktop session through Microsoft's RDP protocol is not enough anymore and organizations (and users) are now demanding remote access with local desktop like experience.

Vendors like VMware, Citrix, Microsoft, Quest, and Virtual Bridges amongst others are in a frenetic race to bite as much as they can from this market. This is a market that will reach 50 million users by 2013 and generate $65.7 billion in revenue according to a Gartner research from 2009.

I gathered some of my thoughts about the subject to compose this article. The essential idea here is to provide an insight of where desktop virtualization is heading to, or should be heading to in my opinion.

The Apps
Many organizations have already understood that VDI is all about applications and personalisation. Apps and personalisation are resultants of what the user is and what he does. Moving forward I foresee VDI solutions allowing GuestOS to be a mere conduit to publish apps. Users and administrators will be able lift the application layer and change the base GOS. The apps will run on top of any flavour of GOS – Windows, Linux, Chrome or Android. What will allow this to happen is application virtualization or application layering.

GOS Parenting

because there's need to support a wide range of hardware's and device drivers; GOS are heavy and contain a large footprint. In the future lighter GOS with smaller footprints and ability to maintain a single master base image and serving multiple instances will be an essential component to alleviate CPU, memory and storage workloads. Today's techniques, such as Linked Clones, are available out of the GOS. I see that ultimately as a feature provided by the GOS. We will see virtual desktops making function and API calls to the base image for those non-cached functions and/or transactions.


A new technique to manage application deployment called layering has recently been released by Unidesk. This technology allows the GOS and the applications to be broken into layers that can be either assigned by the administrator or installed by the users. The technology still maturing but it provides the administrator with the ability to control what application layers users are allowed to have on their desktops environment. Considering that applications are now layers (imagine something like sequential ESX snapshots) they should also be self contained and GOS independent. This will allow application portability across GOS (The Apps) and multiple end-point device support, such as MAC, iPhone, Sius, Blackberries or your TV.

Public VDI Brokers

Today VDI brokering is an organization's internal function. I envisage the existence of public VDI brokers with standardised set of security frameworks that will act as the secure gateway to desktops inside organizations. There will be no requirement for VPNs from an end-point device to the virtual desktop and all devices will have a factory default application for the public VDI brokers. This public multi-tenant VDI broker will allow seamless user experience from multiple end-point devices from anywhere, anytime, and will also be able to broker to a number of VDI infrastructures.

The Cloud
Desktops hosted off-premise, on the cloud, does make sense to me; however bandwidth and latency are still a problem to be addressed. Using public VDI brokers mentioned before users will be able connect to public non-persistent desktops on the cloud whilst secure mechanisms will guarantee that user profiles are safely download from the organizational premises.
User profiles and personalization would reside inside private networks being able to be applied to any public desktop, with any GOS, on the cloud. This once again brings us the idea of application and profile independency. As of today most of the security mechanisms are based around providing access desktops. Tomorrow I see major security mechanisms inside the desktops to protect the user data.
The idea here is cloud-enabled public non-persistent desktops!

Thin Desktop (Cached Mode)
Thin Desktops ate for portable devices. They are the same desktop, however will incorporate a feature to allow synchronization of critical and selected user data with end-point devices. Bookmarks, applications, folders and files could be bookmarked for offline use with iPAD or Mobiles as an example.
These devices will have been shipped with runtime plug-ins to allow the download of those virtualized applications, and they will be able to execute the synchronized offline content.

Online Application Stores

Online application stores will provide application on-demand. These applications will have a standard common factor compatible with most GOS or devices, which will have a plug-in pre-installed to support these stores.
The administrator will have the ability to pre-select or assign applications to be used by users. These applications once executed at a desktop will have its execution linked to the user's profile and will be available for automatic download from any new desktop the user connect to.

A first step has been given by Worth checking out!

User Profile Autonomy
The user profile holds information about the GOS user preferences and application customizations. Moving forward I see these same profiles hosting information about integration with the public application stores, list of applications allowed and blocked etc. User profiles will not be only compatible to a single GOS, but portable to multiple GOS and end-point devices. I also predict the concept of public or cloud based profile and user personalization datastores, hosted by companies like VeriSign and RSA in partnership with VMware (RTO) or AppSense.

In a simplistic approach the traditional VDI architecture looks similar to this:

I am idealizing something more in line to this:

The Client Hypervisor
This is subject for a long discussion but let me cut the crap and go straight to what I think. Type 1 hypervisors will without doubt change the way organizations think about employee's owned computer.

I personally like the idea of the organization's owned asset. At the end of the day laptops are a work tool and I never had to buy my own laptop to work for any organization. On the other side, employees don't want to have their laptop prepared by their company to accept two or more different GOS.

Unless client hypervisor comes integrated with the hardware I don't think there will much space for adoption. Soon or later hardware vendors will ship their hardware with Type 1 hypervisors, but even like that I have my doubts about adoption because that would defeat the purpose of the whole VDI stack.

I consider the ability to port profile and personalization to multiple environments more important than being able to run a local GOS secured by the organization's policies. Remote Desktops will be available to anyone, anytime, anywhere.

Extreme cases like when there is no internet connectivity might be a use case for Type 1, but then again, there will be the offline and cached modes to allow the user to get to the critical data.

Display Protocols
There is a big argument about display protocols going on at the moment. I'll not comment on performance or quality of each one of the products; however I'm confident that existing and upcoming technologies will be able to deliver the desktop like experience. Some performance improvements and features implementation are still needed but as times go by they will all be able to deliver on that. This is probably one of the VDI technologies that will evolve faster than any other.

These are some of my thoughts about the future of VDI and I think everything I list here is achievable with existing technologies. What are your thoughts on the future of VDI?


No comments: