Virtual desktops and thin client computing

The Wall Street Journal had a story yesterday called Businesses Take Another Look at Virtual Desktops*.  Being that it's the Wall Street Journal, the article focuses on the news and financial aspects of the story. This is a techie's perspective on the topic.

The article says that "The so-called thin-client revolution has been touted before but has so far failed to arrive" and then goes on to discuss companies converting to virtual desktops and the advantages of doing so. No mention is made of why thin-client computing remains a small niche. There are reasons.  
 
Let's start with money. According to the article, thin client computing saves money in a number of ways.

For one thing, companies don't have to replace PCs so often in the future. This seems a bit vague to me and it's probably true with or without thin client computing. As computers get faster, the replacement cycles of years past may not hold up now. Most computers are fast enough for most users most of the time. And Windows XP is not going anywhere soon.

Savings are also said to come from eliminating deskside visits by technicians. Is this really that big an expense? Don't IT departments have access to remote control software? 

Electricity must also be expensive, the article discusses savings that "can" go up to 40%. The savings come from thin client devices that are smaller and less powerful than full-blown PCs and don't have internal hard drives. Physical problems with hard drives also go away, another cost savings.

Savings are also said to derive from reducing viruses and security violations. Exactly how is not mentioned and, frankly, it's not obvious to me. Thin clients are more a hardware thing than a software thing. That is, they offer a different desktop OS delivery mechanism, which is apples and oranges compared to viruses.

To put this in perspective, the savings don't come up-front, they are promised for later. In fact, the up-front cost of desktop-resident thin client devices is roughly that of PCs, from $200 to $1,000 per user according to the article.

As a netbook fan, I can't help but compare that to the cost of new netbooks, many of which can be had for under $400. Someone upgrading a desktop computer, can connect their existing mouse, keyboard and monitor to a netbook and still have the advantage of mobile computing if need be. But, I digress.

Potential cost savings are balanced out by something not mentioned in the article, thin client computing adds costs and complexity. And complexity has a way of adding unexpected costs down the road.

The obvious costs are:

• You need expensive high end servers to run multiple OS instances and all their applications
• You need to pay for software to run on these servers that enables them to concurrently run multiple OS instances. VMware didn't become the company it is by giving away software.
• You need in-house technical expertise in the new environment.  
• You may need to upgrade the networking infrastructure to handle the increased load
• You may need increased security for the high end servers that run everything

WHY THIN CLIENTS

Leaving costs aside, why would a company want to convert to thin client computing?

One cited advantage is centrally installing software updates instead of working for several hours over the weekend. Is this really a big deal? Wouldn't a reasonably large organization already have software that distributes updates to their computers? There are many such products, some can even turn on computers that are off so that updates can be installed in the middle of the night.

One upside not mentioned in the article is the fact that without a network connection to a server, the thin client devices are useless. Thus, no one is going to steal them. Bad guys steal stuff they can use or sell. There is no stand-alone use for these things and probably not much of a market to sell them into.

AGAINST THIN CLIENTS

Ignoring costs, there are some downsides to thin client computing.

In designing anything computer related, a single point of failure is always a bad thing. Virtual desktops introduce a couple single points of failure: the network and the servers.

When a server fails, many people are impacted. High end implementations will consist of multiple servers and can probably handle some amount of fail-over. Even so, it strikes me as a bit more risky than having everyone with their own computer.

Even if the servers and the network don't fail, they may not be up to handling the load. The IT department may get a new class of complaint, things are running too slow. Just like in the mainframe days.

Then too, there will likely be some computer users for whom virtual desktops are not a good fit. Users that need high end graphics or do sound and/or video editing for example.  

Then there's the issue of working off-line. You can't. The picture accompanying the article shows two people at a computer connected to a small thin client device with tons of wires coming out of it. How that employee works away from the office isn't addressed.  

Thin client computing not only introduces new software on the server side, but also on the client side. The software used in thin client devices varies drastically. Wyse, for example, offers devices running Wyse ThinOS, Windows CE, Linux, Windows XPe and Windows Embedded Standard. Somehow, someway, someday this client-side software will have to be updated; be it for bug fixes, new features or lack of support due to old age.

FLIP SIDE

I'm not against thin client computing. Years back, I worked in the Research and Development department of a large Wall Street company when thin client computing was just emerging. The Citrix product at the time was called WinFrame (as in Windows mainframe) and I was a big enthusiast.

I still remember the thrill of seeing a PC running an old version of Windows use a thin client application to access a remote server running a new version of Windows. Putting the display in full-screen mode offered a sort of a computing Jekyll and Hyde. It was a hard sell though, for the reasons cited above.

Still, there is a place for thin clients. For example, Gordon Haff over at CNET recently painted a picture of a computer user that desperately needs someone else to handle the care and feeding of their operating system for them.

And the Kentucky school district cited at the end of the Wall Street Journal story is an interesting wrinkle. Rather than buying new hardware, they use their old PCs as thin client terminals. All they need to do is boot the machines using software either on a CD or a USB flash drive and totally bypass the operating system installed on the hard drive. In fact, they can remove the hard drive from the old computers.

The decision to move to thin client computing is complex, with many pros and cons that need to be weighed. The Journal article could have done a better job of painting the overall picture.

Update October 3, 2009:  Back in August an article in Computerworld by Natalie Lambert discussed some myths about client virtualization and had this to say about the pros/cons: 

Client virtualization includes four technologies ... Together these technologies provide IT ops professionals better manageability; data and device security; more tools to comply with regulations; and business continuity in the event f disaster, loss, or workforce interruption. However ... vendors often ignore these benefits and make false claims about potential cost savings.

*This article is only available online to paid subscribers of the Wall Street Journal's website.