If you would like to be notified of when Scott Lowe releases the next part in this article series please sign up to our VirtualizationAdmin.com Real Time Article Update newsletter.
If you would like to read the first part in this article series please go to VDI Economics and Reality (Part 1).
Introduction
In part one of this series; I discussed the direct economic challenges associated with VDI implementations. However, there really is more to VDI, which I’ll cover in part three.
When I first started considering the possibility of VDI, I was instantly hooked, but cautious in jumping on the bandwagon. Today, I’m glad that I’ve been cautious. No matter how you look at it, every technology implementation needs to have some kind of ROI, whether it’s a direct return in hard dollars as a result of money saved or whether it’s derived from new opportunities presented through the implementation efforts. We’ve established a very small lab project at Westminster College and, from a pure technology standpoint, VDI is really cool. These days, I’m all about “abstraction” – that is, being provided the ability to “float” workloads on top of changeable, expandable, movable infrastructure. These kinds of efforts – virtualization in general – certainly improve agility and can improve the bottom line; we have only to look at the savings wrought by server virtualization to see this in action.
VDI is a different beast, however. Desktop virtual machines are spun up and torn down much more frequently than server virtual machines. Further, any performance impact is much more apparent to the user. With a file server, for example, a user will forgive slow access for a period of time and may not even notice it. But for general desktop use, performance must be constant. Obviously, both server and desktop-based virtualization workloads need to be well architected and well maintained, but even the best infrastructures suffer problems from time to time.
As I’ve studied and implemented our very basic VMware View environment, I’ve stayed laser-focused on two items: total cost of ownership and the end-user experience; the end user experience generally translates into employee productivity. There are certainly benefits to be had with VDI and I’m absolutely not arguing against them, but without a good business justification (read: ROI) and as close to a guarantee as possible regarding the end user experience, I’ve paused in my VDI zeal, stepped back and really started looking at the “why” behind VDI.
Why?
I initially approached VDI from a very naïve perspective. My early thinking was that an ability to buy cheap endpoints that I replace every seven years must translate into cost savings. Sure some servers would be needed, but I quickly moved away from this naiveté and began taking a more holistic look at the VDI question. This is one reason that I’ve spent so much time exploring the technology rather than simply jumping in and taking off. As is the case with everyone today, budgets are tight and we have to do everything we can to stretch dollars as far as possible. It’s critical that technology not be approached simply for the sake of technology. In the world of IT, business outcomes are king. Although we (that is, Westminster IT) do undertake some of what I would consider “internal IT” projects, the fact is that any improvement in our own processes translates directly into more time to spend on direct business projects. Often, these internal projects also help us streamline our support processes which can translate into quicker turnaround on business requests. So, from that standpoint, I needed to wrap my head around the outcomes side of desktop management more than the input side and make sure that I was considering all of the key variable and, from there, choose a solution that makes sense for us.
As I began to consider how I was approaching VDI, I was attempting to identify ROI on the technology. While I could define the ROI as far as many outcomes were related (see below), I continued to struggle with the economics (TCO), particularly as it relates to our desktop goals. Unfortunately, like it or not, there needs to be a balance between hard ROI and TCO.
Our goal is to implement a desktop infrastructure that:
Is secure
- VDI
§ Everything stays in the data center; very secure and resilient in the face of hardware failure due to well-architected redundancy
§ Mobile users can be required to connect to their VDI-hosted desktop in order to maintain the centralization of the desktop environment - Traditional desktops
§ If people save locally, all bets are off as far as backup and recovery are concerned|
§ Mobile users may still have data that goes off-site
Enables mobility
- VDI
§ Run the exact same desktop from any device from anywhere
§ If vendor licensing permits (i.e. SPSS), students could conceivably connect to a virtual desktop to make use of campus resources, which enables all kinds of new possibilities, including creation of “on demand” labs in which students bring their own laptops and use a consistent, common campus-provided desktop - Traditional desktop
§ Desktop is likely not available from off-site
§ May require additional software, such as a LogMeIn subscription or implementation of Terminal Services infrastructure, including a gateway and multiple application servers
§ For students, access to campus licensed software requires a visit to a physical computer lab
Does not require constant replacement
- VDI
§ Server-side infrastructure undergoes continual replacement and augmentation as necessary
§ Client side lifecycle can be extended - Traditional desktops
§ Are typically on a 3 or 4 year replacement cycle
Can be easily managed
- VDI
§ Management tools are well-tuned to the needs of VDI and the central focus lends itself well to easy management - Traditional desktops
§ Unless specific tools are used, each PC can be an island
At first glance, it may appear that VDI is far and away the best choice in every one of these scenarios, but let’s look at some ways that the traditional desktop model might be augmented in other ways to achieve similar benefits at a reasonable cost.
Security
- Enforce policies redirecting saved files to backed-up network locations. I can’t tell you how often I’ve seen people save documents locally for “security” or because they’re afraid that the network will go down. Simply put, today’s well-designed networks simply work. Sure, there will be some downtime every so often, but the benefits of saving files to a central location far outweigh the detriments. In this case, organizations that have well-established document management policies will be served very well.
- Require the use of full-disk encryption on mobile devices. Windows 7 includes BitLocker. Truecrypt is free. Make it happen!
- Consider the use of Terminal Services for application access – particularly sensitive applications – as an alternative to VDI or running applications on mobile devices.
Mobility
- Again, Terminal Services might fit the bill here. Users can use some of the new features in Terminal Services – such as RemoteApp – that enable a highly-mobile environment.
- When coupled with appropriate security mechanisms, this kind of architecture can provide a great balance between traditional desktops and provide additional capabilities.
Replacement
- This is one area where I’m not sure the traditional desktop model will every really beat a terminal-based VDI scenario. There are simply physical issues that result in many terminals holding up for much longer than desktops. For example, most terminals have no moving parts. No moving parts generally equates to a much longer equipment lifecycle, particularly when the endpoint workload doesn’t need to be considered. When it comes to terminals, the endpoint workload consists of rendering the results of an RDP, ICA or PCoIP connection, not handling the actual processing.
- On the desktop workload front, implementing a split model – that is, having the base OS run on the endpoint and using something likely RemoteApp to run the applications, you will achieve something similar to the way that VDI operates, at least from an application perspective. So, if you have super-reliable hardware and you can shift the processing burden to heftier servers, it might be possible to extend the desktop replacement cycle a bit.
- For us, let’s also consider the student use case. No matter what, our students are going to bring their own PCs to campus. We could integrate their machines with or VDI infrastructure, but we’re still investigating what that means with regard to VDA licensing and other application licensing. If we need to get VDA licenses for all of our students, I can safely say that there will never be ROI and it would be simply unaffordable. Further, some of our software vendors make it difficult to provide access to their software in a way that provides the kind of flexibility we’re looking for. Even some vendors that provide concurrent-type licenses balk when it comes to providing access to their software in a virtual environment. I’m certain that we’ll see yet more costs from some of these vendors as they attempt to find ways to wring more money from their customers that need to use software in ways that weren’t possible just a few years ago. In short, when it comes to computer replacement, we don’t need to worry about student computers, but as we start to consider elimination of physical labs by implementing virtual ones, we need to keep these machines in mind.
- No matter what, sticking with traditional desktops means sticking with a replacement cycle.
Management
- It is possible to completely manage a traditional desktop environment.
- At Westminster College, I will be the first to admit that our desktop environment has been only loosely managed over the years and that we’ve been more reactive than proactive when it comes to management. We’ve used WSUS for centralized patch management and made primitive use of Group Policy, but that’s about the extent of our intentional desktop management efforts.
- We’re in the final stages of implementing System Center Configuration Manager 2007 to provide software deployment, centralized patch management, reporting and inventory. SCCM 2007 is a cost-effective way for us to handle the desktop management task and will be even better once Microsoft releases the 2012 version of the product.
- Again, a RemoteApp/Terminal Services-based infrastructure can help here, too, particularly with regard to software deployment.
Summary
I want to make it absolutely clear that I’m far from anti-VDI; nothing could be further from the truth. However, in attacking the problem from a CIO perspective as well as that of a technology lover, I’ve found it invaluable to take a higher level look to make sure that the implemented architecture meets as many needs as possible in the most cost-effective way possible. I tend to question myself a lot up until I’ve made a decision at which point it’s go time. In many cases, VDI will be absolutely the right answer while in others, it won’t. It’s clear that there are still a number of other benefits to VDI that may never be captured by the traditional desktop model and those are the items that I’ll attack in part 3 of this series.
If you would like to be notified of when Scott Lowe releases the next part in this article series please sign up to our VirtualizationAdmin.com Real Time Article Update newsletter.
If you would like to read the first part in this article series please go to VDI Economics and Reality (Part 1).
