Wilco van Bragt - LinkeIn Wilco van Bragt - Twitter rssa 

Back to the future: PDI

During the E2E Virtualization Congress London in November 2011 I gave a presentation with this title. The presentation was based on an opinion article (in Dutch) I wrote in March 2011 for the website of the magazine computable with the title the Traditional Desktop is going to disappear.

This article will combine both the Dutch article and the presentation into a new article about the topic that a Physical Desktop is still a good possibility to offer the applications to the end user (or combined with other techniques).




The Fat Client is one of the most used devices for years to offer access to applications and data to the end-user. Many ICT departments found this desktop difficult to maintain and labor intensive.  The alternative was a Service Based Computing infrastructure, but also this technique had some disadvantages like limited usage rights, difficult to install application, vendor support issues and slow performance. The SBC vendors are working hard both on optimizing as developing new technologies. Examples are the introduction of VDI (Virtual Desktop Infrastructure) products and optimization of the used protocols like Citrix HDX, Microsoft RemoteFX and VMware PcoIP. Also more and more vendors understand SBC and VDI and are supporting their application on such technology.

With all these developments it seems that a just a matter of time that the traditional desktop will be going to disappear. But is that actual the case? Beside all these technical improvements on the SBC/VDI area (the backend) several others factors play in the part.  The first aspects are the software and hardware requirements, especially for VDI environments. Think about hardware for hypervisors, (management) software for the hypervisor, licenses of the VDI software, Microsoft licenses required for running virtual machines and diskspace on the SAN. The second aspect is the knowledge of the administrator of the environment. A VDI infrastructure requires specific trained and skilled personal; a "normal" Front Office employee does not have this qualification. SBC/VDI implementations are causing that the former system administrators will have more activities that were done by Front Office employees in the traditional desktop infrastructure. But actually the last aspect is maybe the most important.  To facilitate these new SBC/VDI techniques also the client should have more capabilities. A simple and light Thin Client is not capable for these duties. Thin Client are equipped with more and better hardware, causing that these devices are getting more.........expensive. As a side effect the management of the Thin Clients is getting more complicated, also because a Windows based OS is required on a Thin Client to facilitate all these new techniques. There are several comparisons and presentations available which includes that there several scenarios in which it's cheaper to use a Fat Client based machine (or minimal the same costs) instead of Thin Client to connect to the SBC/VDI infrastructure.

Every management will ask the question in which way the SBC/VDI environment will lead to lower cost or at a minimal what the added value is of a SBC/VDI implementation. In many cases only the management aspect is the actual used reason (difficult to maintain, inflexible and unmanageable).  However if we take a look how SBC/VDI infrastructures several technique and concepts are being used, why should we not use these techniques on the Fat Client. The hardware is not the actual issue, but the way the Fat Client is installed, configured and management. By applying the layering principle and using the same concepts we can gain the same results on the Physical Desktop. The layering principle is based on splitting up the several components, so these will become (mostly) independent of each other.

There are several articles about the layering principle, where there are different layers available. I will use the most important layers.

  • Operating System Layer

 Rolling out operating systems is no "rocket science" anymore, both available via paid software (Altiris/SCCM/and others) as by Microsoft provided standard functionalitye (WDS) for installation and maintenance via products like WSUS or third party alternatives. New techniques like OS (via Citrix Provisioning Services) are arranging that this layer is more efficient to manage and roll-out by on-demand streaming of the OS image to the workstation. One step further are the developments on the client hypervisor type 1, now available by a few vendors like Citrix XenClient, MokaFive and NxTop and the upcoming new version on Windows now labeled as Windows 8 which will includes Hyper-V 3.0. The technique of client hypervisor is based on the same principles as the server hypervisor solutions, so it's not completely new. However there are some challenges, mostly in supporting the big diversity of client hardware and delivering the hypervisor and image to the client. However I think that using a single image (based on a client hypervisor) for all workstations could be possible in the (near) future.

  •  (User) Application layer

 This second layer is where it's all about today. Delivering the applications (and corresponding data) to the user, so they can perform their daily activities. The traditional installation of applications, maintaining applications and installing application next to each other on one workstations are combined the largest bottleneck of the traditional desktop. Via Application Virtualization these challenges are (mostly) solved and this technique can be called proven technology nowadays. Also webbased applications and application hosted in the cloud are possibilities that will be available as a future alternative.

  •  User environment layer

 The last but not least layer is the user environment layer. This layer contains of all user configuration (applications, windos), user data, but also the assignment of application shortcuts, data folders, printers and similar. 

Currently redirecting the data folders of the end users or synchronization for laptop users is the best available options. Although folder redirection has disadvantages, there is currently no better solution available. For other settings/topics User Managements Products are really advises. In my opinion a User Management product is essential in current implementations. Examples are RES Workspace Manager, AppSense Workspace Management, Flex Profile Kit, VUEM or for basic implementations via Microsoft Policy Group Preferences.

This layering is a technical solution to apply the most important concept. A uniform and standard arranged workstation, which can be used by any end-user in the company to their daily activities. In other words a workstation is no longer unique and a disturbance on one workstation does not cause that a single user is disturbed in his activities. The user can continue his work on another workstation, while ICT "solves" the issue by exchanging the workstation or redeploy it quickly. In the case we are using OS Streaming, application virtualization and user management software a simple restart is enough to get the machine in a good state. So will the traditional desktop going to disappear? Definitely, the desktop will be going to disappear. Not the hardware what many people are thinking, but the method of installing and maintenance the desktop will change. I gave this vision/concept the name PDI, which stands for Physical Desktop Infrastructure.

Is this PDI concept the solutions for all issues? The question is pretty easy, namely No. In which way you need to offer remote access to the environment? Does it fit in scenarios where there are many small branch offices with small bandwidths? In such cases there are some challenges using the PDI concept without comprises. Logically there are solutions for that, for example using a Server Based Computing solution in these scenarios.

And this brings us to the real essence of this article. Currently there are several solutions available offering a desktop to the end-users, all with their own characteristics, advantages and challenges. In my opinion VDI is currently overhyped and too much implemented while other techniques probably suits better. Actual no of the current options can offer a 100% ratio, so combining these techniques is still required. PDI can be used perfectly together next to SBC or VDI infrastructure; actually I dare to say that PDI is required next to the SBC or VDI infrastructure to keep the desktop across the platforms uniform both from an end-user perspective as a maintenance aspect.

Summarized I try to tell you in this article that a Fat Client is not bad at all, when it is well-thought installed, configured and maintained. It's all about the management and concept used that creates a good running infrastructure however this is SBC, VDI, PDI or a combination of these. Really consider a PDI infrastructure in the first phase as an option for the new desktop model (next to SBC and/or VDI) is created for the organization. PDI may fit perfectly in the organization or it can be combined with a SBC or VDI solution for those end-users that does not fit in that model (left-overs).