Virtual Desktop Infrastructure (VDI) is poised to change the face of client support on enterprise networks. We talked to Virtual Bridges CTO Leonardo E. Reiter about VDI and the promise of the virtual desktop.

Q: Let me start by asking, what is VDI?

A: In a nutshell, VDI refers to hosting and consolidating virtual desktop workloads and delivering them to clients. It stands for “Virtual Desktop Infrastructure,” a term coined to describe this process. About 10 years ago we started delivering these types of solutions to customers looking to cut desktop management costs and improve efficiency.

It helps to think of VDI solutions as “generation 1” and “generation 2.” Generation 1 was about hosting virtual desktops on servers and delivering the user experience to connected clients via some display protocol. This is very straightforward but is a concept that is now 10+ years old. The majority of VDI solutions on the market today still fall into this category.
Generation 2 (or Gen2 for short) takes the basic principles of Generation 1 but addresses the major gaps between “promise” and “performance,” such as unreasonable server and storage requirements, overwhelming complexity, lack of coverage of use cases, such as disconnected use, and even Linux virtual desktops. This is what we do today and what is driving this technology into the mainstream.

Q: The thin client paradigm has been around for years. How does the modern concept of VDI differ from other thin-client terminal service scenarios discussed in the past?

A: Thin clients trace their roots back to dumb terminals used to access large monolithic systems such as mainframes. They represent the pinnacle of access efficiency because there is little or no state on the device. Unfortunately, they do not cover all use cases that modern organizations look for out of computer systems. In short, if the user is not permanently attached to a network or always sitting at his or her desk, the thin client model falls apart. Also, if you make thin clients too “dumb,” they do not provide enough functionality to deliver new types of content such as rich multimedia. If you add client-side software to achieve this, then you re-introduce management headaches into these endpoints that are supposed to require little or no management. While thin clients are an important element of mainstream VDI, they are not always the end-all solution.

In terms of the other end of the infrastructure, VDI differs greatly from terminal services or other legacy server-based computing solutions. The PC let the “genie out of the bottle” in terms of flexibility and user personalization of desktops. Users feel a natural sense of entitlement to have custom wallpaper, private documents, and, in some cases, custom applications as part of their unified desktop experience. In a published desktop or terminal server environment, users have far less control over these attributes and, on top of that, since they share a kernel with many other users, are at the mercy of the overall health of this shared system. For example, a single user in a terminal server environment has the ability to crash an entire shared computer, which is simply not possible in VDI because each desktop workload is segregated in its own isolated virtual machine.

Another fundamental difference is application compatibility, both between applications and the terminal server as well as among different applications sharing the same space. This is especially problematic in the Windows world since most ISVs do not even support their software on Microsoft’s server operating systems (such as Windows 2008), only desktop systems such as Windows XP and Windows 7. So if you want to serve many (even most) Windows applications in a terminal server environment, you are basically on your own when it comes to vendor support. Then there is this notion of “food chain dependencies” – different applications require different versions of libraries and even system configurations. This leads to a phenomenon called “server bloat,” where terminal servers are often underutilized and more must be deployed just to deal with more applications. This was the original “killer app” for server virtualization – hosting Citrix severs on VMware infrastructure to improve utilization. Obviously this is a problem that is non-existent in VDI because it is possible (and in fact trivial) to mix and match workloads on the same physical hardware since each runs in its own isolated space.

However, there is one more big problem with terminal services, and this is really the “Achilles heal” ... users must be connected full time to the infrastructure to use the functionality. This is obviously not how many users work today, so this turns terminal services into a point solution that works for some use cases but in no way can replace ordinary desktops across the board. In fact, this is one of the pillars of VDI Gen2 – the ability to support both connected and disconnected use from the same infrastructure and yet retain all the benefits and efficiency of centralized management.

Q: What are the benefits of VDI over conventional approaches to centralized desktop management?

A: There are three major benefits that I can summarize as follows:

1. Transforming desktop TCO [total cost of ownership] – The ability to centrally manage desktops in a 1 to many way reduces the cost and complexity of applying patches and changes to desktops individually, even if aided with management software. Any time you have state on an endpoint, it becomes a liability. It is okay to have transient state on it because this does not impact anything if the device is lost, stolen, or fails. But storing permanent state on a PC, even if centrally managed, is what leads to astronomical desktop TCO most organizations face – in some cases several times the cost of the PC, year after year.
2. Improving security – By taking sensitive data off the end point, organizations can store it “behind the keypad” in the data center where it is protected, backed up, and highly available.
3. Promoting agility and flexibility – The notion that people sit at their desks all day and use the same computer to do their work, day after day, is completely outdated. In any company, large or small, employees may use home computers, thin clients, tablets, disconnected laptops – sometimes all in the same day. Their desktop experience must be consistent, secure, and highly available across all these use cases.

Q: What do you need to make this work? Is every corporate network ready for this kind of solution, or are there some minimum hardware and bandwidth requirements that need attention?

A: Traditionally VDI has been an extremely expensive, complicated project requiring major physical infrastructure upgrades before you could start realizing the benefits of the software. While networking is a key element, storage is an even bigger one. With Gen2 solutions, the burden is far less, since network topology is less critical and storage capabilities are far more reasonable due to platform efficiency.

Of course, the exact answer to the question is “it depends.” This is why any successful VDI project, no matter who the vendor, must begin with a thorough assessment that includes inventory of existing resources, use cases, and process redundancy. From there, organizations must decide what to tackle first and come up with a realistic approach to the migration. The key to the assessment is honest self-inspection, as well as weighing goals against available budgets. Organizations should be prepared to pilot solutions and establish clear success criteria, while taking a phased approach to this upgrade.

Q: Why isn’t everyone doing this? What are the pitfalls that admins and users encounter when implementing VDI technology, and what are your thoughts on what to do about them?

A: Simply stated, it’s because VDI Gen1 has failed to deliver on the promise of the technology. When we had our Gen1 solution 10 years ago, it was sufficient for use cases at that time, but not for the mass market, and certainly not for the use cases of today – so not long after, we started working on Gen2. When other vendors got into the space with their Gen1 solutions, they came from different directions than desktop virtualization: one major vendor from server virtualization and the other from application delivery. To get from there to VDI for these players happened via technology acquisitions and third-party “ecosystems,” resulting in additional cost and complexity.

Then there’s the killer: again, lack of coverage. Solutions that are data center bound are simply outdated. Yes, we fully believe in data centers, but you cannot rely on the data center for 100% of all operations in modern deployments. If a user gets on a plane, they are no longer attached to the data center. If they log in from Starbucks, unless they VPN in and suffer the performance hit to their remote access, they are not connected to the data center. Finally, if the data center is in Texas and the user is in India, relying on a constant connection to the data center is asking the user to accept a less than optimal experience when using remote applications. Even if the bandwidth is good, the latency is typically unacceptable, and the uptime is not always guaranteed across oceans. The decentralization of VDI is key to widespread adoption, whether it is to serve disconnected users in a reasonable and secure way or distribute infrastructure to better respond to user needs. VDI Gen1 simply falls flat here, which is where Gen2 picks up.

Also, Linux support has also been a problem for many VDI solutions (although not for ours). Almost every large company has at least some Linux desktops, and certainly public sector, government, and higher education do as well. Organizations looking to standardize on VDI typically have to face the fact that some vendors simply don’t support Linux – either because they are so large that the business case is not there, or because they have delicate politics with partners such as Microsoft.

Finally, there is the new trend of utilizing service providers to host at least part of the infrastructure – essentially, utilizing the “cloud.” Some vendors simply don’t have an adequate solution to give to service providers, much less provide a hybrid model where some processing takes place on premise, and some in the cloud. This all comes back to the decentralization of VDI and why Gen2 is picking up where Gen1 left off.

Q: What is the minimum-size network that might want to consider VDI technology?

A: The beauty of a flexible solution is to be able to scale down as well as scale up. In our model, we can support as few as 5-10 users or as many as 1,000,000+, with the same exact product suite. There are certainly cases where an office of 10 users would benefit from a small VDI deployment. Desktop management is an issue for any organization, whether they deal with it formally or not. If employees are busy applying patches or reinstalling software, they are not being productive. VDI can make IT management for small organizations more efficient, even if no other desktop lifecycle management processes are in place.

Q: So I guess what a CIO or other IT manager would want to ask is, does this really save money? Often that question comes down to saving time. Can you support more users with the same IT staff in a VDI environment? If so, why?

A: Absolutely. But you have to use the right technology, be methodical about how you apply it, and be honest about what you are trying to achieve, and in what timeframe. A well-deployed VDI solution that leverages single-image management allows organizations to be proactive with IT resources rather than reactive. Rather than fielding support calls about blue screens and running to users’ desks to replace equipment, IT resources can be rolling out the next generation of systems that are core to the company’s business. I have personally witnessed customers reduce their help desk activity by an order of magnitude, freeing up IT staff to focus on other pressing needs. Reinstalling Windows, cleaning computers of viruses, and replacing hard disks, for example, is typically not the best way to utilize IT resources. VDI can help, especially a Gen2 solution.