Virtual Desktop Infrastructure isn’t a throwback to dumb terminals


The server virtualization craze has huge momentum and popularity as we all know. It’s pretty obvious that the reason for this is economics. The utilization of capital assets, specifically servers, is far better after virtualization than it is before virtualization. So developing an ROI that shows very fast payback is not hard to do, and hence it is popular.

The natural inclination is to view everything painted with the “virtual” brush positively; the word virtual makes people to assume that VDI brings the advantages of server virtualization to the desktop. Make sense?

Server virtualization and VDI are not the same. I am not saying VDI is bad, but the ROI and metrics used to judge VDI must be different than those justify server virtualization. I have seen many customers cancel suitable VDI projects after the design phase showed that VDI would fail to meet the customers’ own criteria; a criteria borrowed from server virtualization projects. Virtualizing a server and a desktop are two very different things. One could argue that the light load on a client or PC provides an opportunity to save money by moving the compute power to a centralized machine.

Then each user needs just a keyboard, display, and communication device; Just that? If you look at a pie chart of the cost of a desktop machine, removing the hard core computational pieces, and in some cases a local disk drive, doesn’t remove much cost. As an example, a popular thin client VDI terminal, the Sun Ray 3i with an integrated 21” display goes for $699. I noticed that you can buy a really nice SONY desktop computer with a 23” integrated flat panel display, dual core 2.5 Ghz. 64 bit processor with 4 GB of DRAM for $799. A whopping $100 savings per seat!

In many VDI projects nobody actually loses their existing desktop machine. Instead, IT delivers centralized apps to the users. This hybrid approach has some great Capex advantages, but compromises the desktop support, security, and availability benefits assumed under a purely virtualized desktop.

After dealing with the complications that arise in VDI deployment, like whose desktop is virtualized, and what applications existed on that desktop that will be delivered centrally, the requirements for the network need to be taken into account.

For example, email stresses bandwidth and latency much less than VDI desktop graphics, so it may turn out that a centralized graphics support computing model simply shifts costs to the network. 1 Gbit to every VDI terminal vs 100 Mbs to every “fat client” email PC? Or perhaps cost effective 11g wireless to the desktop PC’s, which all include a Wi-Fi radio. Works great for email and web browsing, but not for bitmapped VDI graphics.

When teams discover the added costs of VDI projects, they emphasize the benefits for VDI that save money indirectly. For example, version control, document control, data security, support scalability and potentially system availability can all benefit with VDI. The economic benefits of these attributes are harder to quantify, and hence harder to justify than those of server virtualization. For some environments, these real benefits of VDI will win the day.

If people can hark back to the “old days” of centralized computing, they might just remember how much fun it was to get to client-server computing. VDI is a claw-back to the centralized model and promises to also claw-back some of the advantages of centralized computing. Keep in mind that CPUs cost more relative to screens, keyboards, and wires back in the day. Now you can’t even build something as dumb as an old IBM 3278 terminal– your toaster has more compute power than it did.

The cost structure of centralization gives you the opportunity to store your users’ data more reliably. SSD pays off in VDI environments for performance, especially to cope with start-of-day boot storms. After all, a slow VDI implementation will make IT customers view the whole project as dumb as the terminals we had in 1978.

This blog is not meant to give you all the answers; else it would be at least as voluminous as an issue of the Economist. VDI projects require careful thought, analysis, and planning. While desktop virtualization offers benefits, don’t rely on a straightforward application of the ROI models from the VMWare project you just did. If that is why you are undertaking the VDI project, you will probably stop before you are done.

So the answer to the question then is no, VDI isn’t a throwback to dumb terminals. Besides, does anyone remember how much those “dumb” terminal cost? Not sure the “old days” were cheaper anyway.

Mike Workman is Chairman & CEO of Pillar Data Systems. Mike has spent his career breaking new technical ground in the storage industry. In his 25+ years in the storage business, Mike's appointments have included vice president of worldwide development for IBM's storage technology division, senior vice president and CTO of Conner Peripherals, and vice president of OEM storage subsystems for IBM. He has a PhD and Masters from Stanford, a Bachelors degree from Berkeley and holds over fifteen technology patents.

  • Come off it Mike, this is drivel and you know it is.

    “a popular thin client VDI terminal, the Sun Ray 3i”

    the Sun Ray is about as popular as herpes.

    ” 1 Gbit to every VDI terminal vs 100 Mbs to every “fat client” email PC?”

    Name one client remote display protocol that requires 1 Gb, you can’t, there isn’t one. Even the most excessive remote display protocols seldom require more than 1 Mb/s.

    “Or perhaps cost effective 11g wireless to the desktop PC’s, which all include a Wi-Fi radio. Works great for email and web browsing, but not for bitmapped VDI graphics.”

    Maybe you should actually try using the technology before criticizing it, there are many people in the real world today using VDI solutions quite successfully on 3G data networks.

    Please, stick to writing about storage As you rightly point out, if you fail to get the storage infrastructure correct for a virtual desktop deployment you will deliver dreadful performance and there are many innovative storage solutions being delivered by startups that are well worth talking about In the context of desktop virtualization.