A good Virtual Desktop Design architect needs to ask, “what is the best solution for this environment?”  To answer that question, an effective design considers – “does the solution meet the users’ needs?

If your virtual desktop design starts with sizing hardware to support the amount of current physical desktops, starts with considering what applications can be virtualised, it has likely started in the wrong place.

In this article we discuss the differences between virtual desktop model choices, and how you can map user requirements to those choices.

In our article on Sorting out Desktop Virtualization we provided categories of desktops that can be provided to users. These were:

  • Presentation Virtualization (PV): run multiple sessions on a server instance – typically on Microsoft’s  Remote Data Services (RDS), (formerly Terminal Services) (e.g. Citrix XenApp, Ericom PowerTerm Webconnect, Quest vWorkspace).
  • Server Hosted Virtual Desktops (SHVD): hosting a desktop operating system within a virtual machine running on a centralized server – also known as  Virtual Desktop Infrastructure (VDI)  (e.g., Citrix XenDesktop, Ericom PowerTerm Webconnect, Quest vWorkspace, VMware View)
  • Client Hosted Virtual Desktops (CHVD):  rather than run a virtual desktop on a server hypervisor, utilise virtualisation to manage the image to be run on the user’s device (e.g. MokaFive, RingCube, VirtualComputer, Wanova)
  • Physical Desktops (PDT):  PC’s, Blades-PCs, Laptops, typically hosting a Windows flavour of OS and installed applications.

Of these – two are ‘classic’ virtual desktop delivery mechanisms, and there is a new “third way”:

  1. Presentation Virtualization (PV) – or Hosted Shared Desktops: Many users to one desktop. Users get a desktop interface, which can look like Windows 7 if that is your wish. That desktop is hosted on a server operating system, and that operating system is being shared by every user on the server. Although restrictions and redirections can be configured to minimise users sessions negatively impacting on other users, there is still a risk of one user adversely impacting others.
  2. Server Hosted Virtual Desktops (SHVD): one user to one desktop. A desktop OS (Microsoft Windows/XP, a Linux variant) running as a virtual machine where a single user connects to it remotely. One user’s desktop is isolated from another user’s desktop configuration. There are different flavours for the hosted virtual desktop model, for example “pooled” (each instance is generic) or “dedicated” (each instance assigned to a specific user).
  3. Client Hosted Virtual Desktops: centralise the management of the image – but host that image at end-point. One image, many desktops.

When delivering a virtualised desktop, PV can be considered in the first instance. It is the most scalable, the most established: this equates to lower risk, although it is not without significant initial investment. A PV solution would be able to support double the number of users that a SHVD/VDI model with the same hardware. On the other hand, SHVD/VDI offers wider application compatibility than PV, while Client Hosted Virtual Desktops require far less investment in terms of data centre server resource and can be utilised for off-line working.

Unfortunately, the choice of models is not an either/or answer. What we could have done above is to look at the models and thought – “which one supports the most users” or “which one is cheaper?”.  That’s not a good place to start. If we’d considered “which model best fits ours users’ needs”, what we need to deliver will more likely be a combination of models and allow a faster, and larger, return on the investment in the technologies.

What Do Your Users Do?

“Complain and demand” may well be your first answer: funny, likely very true – but let’s focus.  A common starting model is to understand user constituencies. There are a number of typical desktop worker scenarios:

  • Mobile workers – Accesses the network from remote, often public locations. They will use mobile devices (laptops, smart-phones, tablets or perhaps even a Nirvana phone – and these devices may or may not be a corporate device). They need secure, 24×7 access-to and storage-of sensitive data. Mobile workers would typically be executives, salespeople or trainers. Data centre based Virtual Desktops (PV,SHVD) cannot offer a solution for off-line working as they rely on an active connection – this could be accommodated by Client Hosted Virtual Desktops. However, server-based virtual desktops offer access to applications without the need to store data on the end-device: indeed the amount of data sent between the client and the server is small, reducing the need for expensive mobile bandwidth allowing mobile users full and secure access to data intensive applications. Note, not all virtual desktop solutions offer a good user experience over low bandwidth links.
  • Office-based workers will traditionally have used a desktop PC while in the office. They’ll use a broad range of desktop and server applications. Office based workers can work at multiple computers: their roles would include analysts, architects, designers or health professionals. An advantage of Client Hosted Virtual Desktops have over both server based virtual desktops for these users is that you can offer better management of such traditional PC environments at far less cost. That said, if you’re considering reducing the number of desktops and introducing hot-desking, or hotelling where users no longer have a fixed device, a server based virtual desktop offers a greater freedom of options. An issue for PV in this instance can be that some applications do not work, or are not supported, in a server operating system environment although for the majority of applications, this is not the case.
  • Task-based workers, like Office-based workers, will work in the office but they will be limited in the scope of functionality that they require from their workspace. Task based workers tend to have the older, lower specification devices.
  • Home workers – can include office workers who need occasional home use due to weather, a family emergency, or even a natural disaster. As with Contract workers, server based virtual desktops is often the best option deliver applications with minimal change to the user‘s home device. As the session is running on a server in the data centre, the hardware requirements of the user‘s home PC can be extremely modest. The installation of an access client on the home user‘s device is often a simple process that requires little to no technical knowledge. Delivery of access to a usable workspace is therefore quick and efficient – allowing the user to be more productive. Once the client is installed, it is possible to allow secure access to applications and services without corporate data or applications needing to be installed on that device, maintaining security and increasing productivity.
  • Contract workers – are non-corporate employees located remotely often working on non-corporate assets, or outdated computers. Server based virtual desktops allow you to accommodate these users with minimal change to that environment. The installation of a client on the contractor‘s device can allow access to applications and services without copying corporate data or installing applications on that device: reducing need for dedicated terminals and maintaining security. As the session is running on a server in the data centre, the hardware requirements of the end device can be extremely modest. However, Client Hosted Virtual Desktops can also be utilised as long as the CHVD does not require that the exising device is altered.

These constituencies can help you relate users to tasks and requirements – but as we’ve seen each constituency can be served by multiple virtual desktop design types. Lets consider those requirements against virtual desktop types.

Mapping Users to Desktop Virtualization Solutions

If we consider what users do, how they do their tasks can be broken down into six activity types -  it is possible to map virtual desktop strategies to those types.

Type Requirements Summary
Presentation

Virtualization

Server Hosted
Virtual Desktops
Client Hosted

Virtual Desktops

A Users need a small number of applications to work.  The need for modification of these applications is small to non-existent and their access device needs to be inexpensive.
B Users have a core set of applications they require to do their jobs.They must be able to modify system-level settings like environment variables, or install their own applications
C Users focus on content creation utilizing Microsoft Office and Adobe Photoshop. They use a browser to search and manipulate content and graphics on-line.
D Users utilize a applications that consume significant amounts of CPU resources when doing certain activities (video rendering, code compiling)
E Users need to utilize applications while not connected to a network
F Remote users need to access central corporate data securely

This isn’t an exhaustive list. You, for your organisation might be able to introduce new requirements. What it does show, is that it is not always one solution – and that user constituencies are not necessarily directly mapped to one type and one type only.

Now you’ve started, where do you go?

A combination of models may appear more complex and expensive, yet the investment can provide greater benefit than a single model because it increases productivity by supporting what the user needs to do, when they need to do it – be they in the office, at home, or out on the road.  By all means start with a  model that supports most of your users; and this may require a phased approach to implementation. It is likely that a model is already in place – PV is incredibly pervasive, where possible – use existing technologies first it simplifies implementation and reduces the time to delivery.

A number of vendors provide you a single management platform to support both PV and SHVD;  reducing the complexity of the design and infrastructure requirements. These vendors include Ericom, ProPalms, Quest, Systanica, and 2x.  At the time of writing, leading desktop virtualisation vendors Citrix and VMWare cannot deliver such an integrated environment. VMWare offer a solution for hosted virtual desktops (View) and to a lesser extent client hosted virtual desktops (ACE) . Citrix’s XenApp and XenDesktop products have different design architectures, they need to be run as two separate entities  which complicates design and management that said, with the release of their XenClient client hypervisor Citrix support managing client hosted virtual desktops and Citrix’s Flexcast delivery technology is geared towards allowing you to utilise Citrix’s product suite to deliver the right environment for your user: therefore it is likely Citrix’s roadmap will converge what is currently three different products into one. While that happens,  Quest has partnered with Virtual Computer – although we’re yet to see tangible delivery of what could be a similar Flexcastesque offering. It will be interesting to see how Ericom, VMware,  and Wanova write their dance cards.

Convergence is not a ‘futures technology’.  RingCube has taken this ‘mix as required’ model to pioneer a new approach to desktop virtualization called Workspace Virtualization. Ringcube have designed their Workspace Virtualization Engine (WVE) to provide the full Windows desktop experience of a local virtual machine with the performance of application virtualization to deliver a unique desktop virtualization solution that eliminates the barriers to adopting desktop virtualization in the enterprise.

When considering a Virtual Desktop Design a good architect needs to ask “what is the best solution for this environment?”  For many, once you’ve considered the needs of your users, it is a combination of desktop delivery models – some virtual, some physical. Ideally the user is unaware of which model is being delivered to them, they consume that service on an appropriate device, at an appropriate time.  Fundamentally, don’t choose the buzzword as the solution – choose what delivers to your users’ needs best.

Presentation Virtualization

Share this Article:

Share Button
Andrew Wood (144 Posts)

Andrew is a Director of Gilwood CS Ltd, based in the North East of England, which specialises in delivering and optimising server and application virtualisation solutions. With 12 years of experience in developing architectures that deliver server based computing implementations from small-medium size business to global enterprise solutions, his role involves examining emerging technology trends, vendor strategies, development and integration issues, and management best practices.

Connect with Andrew Wood:


Related Posts:

3 comments for “The Virtual Desktop Design Maxim: Start With User Requirements

  1. January 20, 2011 at 9:13 AM

    Andrew,

    I couldn’t agree more with your article. I’d like to add my two cents. You can group your users into categories as you have shown as a great starting point, but actually getting an intimate understanding of how these users work on a daily basis is the critical component to a successful virtual desktop implementation. As we both know desktop virtualization is not a stand alone solution, but will more than likely be a hybrid solution of SBC and “VDI”. By gathering the necessary “in-depth” data about your end user community and “tiering” the user groups as you have shown is the start of building the bridge between vision of what the solution should be and reality of where the infrastructure is today. This gap between vision and reality is what I call the “dragon gap”. This is the place on ancient maps that cartographers drew in serpents and dragons for the unexplored areas of the world. It is the same for IT organizations today and their expeditions to get to successful virtual desktop implementations. There are many unknowns/dragon gaps about the end user environments. Organizations need a toolset that gives them this in-depth data about the end users in order to remove the serpents and dragons from their “map” to virtual desktop solution success. Then they need a solution that lets them continue the monitoring and managing of the crucial metric of “end user experience” and be able to quantify that and report about it.

    Great article mate, looking forward to more from you.

    Cheers
    Michael

  2. January 21, 2011 at 2:21 PM

    Michael,

    As ever, fair points well made.

    The thrust above is to get an understanding of the fact that – some models can be more appropriate than others for different user groups. Quite often one model is considered at the expensive of others. But – don’t focusing on technology, focus on the users. IT should be an enabler for business.

    You’re right – understanding the users can be complex. An early question may be “does this app run in a virtual desktop”, better early questions are “does the user use this app?” “what other apps does the user use that we might not know about?”. Again – focus on the users, not the technology. Centrix Software have some useful tools for this.

    And then of course, once the service is in – support focus traditionally was “is the server up” “is it hot? is it cold?” Not concentrating on the users. Focusing on the technology. A better use of resource would be to monitor the user experience – is the service responsive, are log on times consistent, how long is a user waiting.

    Citrix have done well to integrate tools such as Edgesight for end-user experience monitoring into XenDesktop/Xenapp. That’s not to say Citrix have the only solution – Bernd did an interesting piece last year on Virtual Desktops and End user experience management (http://www.virtualizationpractice.com/blog/?p=4178); maybe its time to re-visit that as I’m sure each of the vendors mentioned there have been working hard over the past year.

  3. Paul Whyton
    January 23, 2011 at 5:56 PM

    Hi Andrew,

    Nicely thought out peice. I have to agree with the explanations you put forward, too often we come up against clients where the first thing they say is “This is what i want” without taking into consideration what their business needs.
    We find that understanding the profile of a user from Application Use, Logon/Logoff, Idle time, movement… is the key to getting the solution right and you are quite correct in that more often than not it is a combination of 2 and sometimes 3 methods of workload virtualisation.

    regards

    Paul

Leave a Reply

Your email address will not be published. Required fields are marked *


six + 5 =