Executive Summary:

This candid account of a Windows Vista migration can help you figure out what to expect in your own environment and how to deal with—or avoid—common mistakes. Typical workstations require a video adapter upgrade, primarily to run the Windows Aero interface, but the cost and effort might not be worth the results. Vista performance was noticeably slower than Windows XP, but Vista SP1 has improved performance. In the end, some of Vista's features such as the improved search engine won over the end users.


I’m not an impulsive person. As a network administrator for a small software company, I tend to look carefully before I leap, and when I started testing the beta releases of Windows Vista, I was even more cautious. We have a lot of serious power users in the quality assurance and technical support departments, and if I forced an OS on them that they hated, I’d never hear the end of it. To begin with, I had to add memory to my test machines and replace the video adapters before I could even run all of Vista’s features. And even then, the performance wasn’t exactly spectacular. But hey, these were betas. Things were sure to shape up by the time the code was released to manufacturing, right?

Well, things did shape up, to some extent, and I went ahead and leapt. I plotted out a migration schedule for my 90-node network that I thought was timely but not impetuous, and converted all of my user workstations to Windows Vista Business over the course of a few months. However, if I knew then what I know now, I would have definitely done quite a few things differently.

A Hard Look at Hardware
The workstations I migrated to Vista were a few years old, but they were hardly relics from a museum, and yet every one required a hardware upgrade to conform to the recommended system requirements for Vista Business. Most of them had 512MB of RAM and needed to be upgraded to at least 1GB; quite a few needed larger, faster hard disks; and every single one needed a new video adapter to replace the basic one integrated into the motherboard—I didn’t have one computer in my whole shop with a Vista-compatible video card.

My users run mostly standard business applications. They don’t need extensive graphics capabilities, and I’m not interested in gaming, so I’m not exactly up on the latest 3D, lightning-fast, water-cooled video adapters out there. In fact, I needed a crash course in the latest video specifications before I even understood the requirements for a Vista-ready adapter. I also had a lot of trouble identifying video adapters that had both the required hardware components and drivers that conformed to the Windows Display Driver Model (WDDM) specification. In those early days, the only way to be absolutely sure a video adapter was Vista-compliant was to install it and run the Windows Vista Upgrade Advisor, a process which can be expensive and lengthy when you’re shopping over the Internet.

I finally located video adapters with the right hardware, the right drivers, and the right price, and installed them in all of my workstations. This whole process complicated the migration enormously, and do you know what? It wasn’t worth the trouble. The video hardware requirements for Vista exist primarily to support the Windows Aero UI. The translucent window frames and Flip 3D interface are reasonably cool-looking, but my users weren’t overly impressed, and I decided that the Aero experience wasn't worth the system resources it consumed.

I ended up switching all of the workstations to the Windows Standard user experience, which eliminates the Aero cosmetics but takes advantage of the WDDM drivers. WDDM is an excellent step forward in driver design. It shifts most of the display driver code from kernel mode to user mode, so a driver problem is far less likely to crash the entire system. However, I could have saved a lot of time and money by leaving the old video adapters in place and running Vista without Aero from the beginning.

Prickly About Performance
I ran the beta version of Vista on several computers, and as is often the case with betas, system performance wasn't great. Of course, when you’re working with a beta, you attribute poor performance primarily to code that hasn't been checked and streamlined for general release. By the time the Vista code was finalized and released to manufacturing, I already had a migration schedule in place and running.

I expected to see a marked improvement in performance when I was running the final Vista code on my production workstations, especially considering that I had installed extra RAM and faster video adapters in them. That performance improvement never materialized, and it still hasn’t, even after installing SP1.

Some people spend a lot of time and money running benchmark tests to gauge the precise performance levels of their computers, down to the last clock cycle. I'm not one of those people. I prefer to evaluate system performance empirically. If I can’t detect a change while performing typical tasks running everyday software, then statistics are meaningless to me.

However, I did run some informal, side-by-side performance tests using two of my standard user workstations, with one running Windows XP and the other running the initial release version of Vista. Overall, the XP machine was palpably faster than the Vista one, even though the Vista computer had more RAM and a faster video adapter. Simple file management operations, such as browsing network shares and copying and deleting files, were particularly slow. This situation would later prove to be a major source of frustration for my users, with many of them asking if they could have XP back.

Although their evidence is strictly anecdotal, several of the department managers in my company believe that productivity has declined since the installation of Vista. Interestingly, it was the managers of the order entry and customer service call centers, two departments that are highly reliant on their computers, who first approached me to complain.

When researching the performance problem, I found some documents claiming that XP file copies only seem faster than Vista copies because XP uses cached I/O differently than Vista does. In other words, they said that XP was returning control to the user faster, but the actual copy process was still going on in the background. This might be true, and it might be an important distinction to programmers and theorists. However, to my users, all that matters is how long they have to wait until they can perform another task. If Vista keeps them waiting an extra sixty seconds while a file copies, and XP lets them continue working, they’ll choose XP every time.

These performance problems have been well publicized, and Microsoft clearly has attempted to address them in the first Vista service pack. SP1 definitely improves system performance somewhat, particularly in the file management tasks that were most irritatingly slow. However, even with SP1, if you were to ask me whether a Vista upgrade improved system performance, the answer would still be no.

The Driver Nosedive
My other big problem was locating Vista drivers for some of the hardware components in my workstations. Some of my network adapters slowed down to a crawl after the migration, and some audio adapters exhibited strange behavior, such as playing sounds from some applications but not others. I’ve been working with Windows since version 3.0, and I don’t recall a new release ever causing driver problems with this many components. I also don’t think it has ever taken this long for the hardware manufacturers to create and release new drivers for their products.

In a few cases, I replaced network interface adapters rather than spending more time fiddling with driver upgrades. Audio drivers are not mission-critical to my users, so I ended up leaving the few problematic ones in their partially operational state.

Would I Do Things Differently?
Being an early adopter was a relatively new experience for me, and I suppose that migrating my workstations to Vista so soon after release wasn't a particularly good idea in my situation. I don’t mean for this article to be a blanket indictment of Vista. When I tested Vista on a state-of-the-art computer with a lot of RAM, fast SATA drives, and a high-powered video card, it performed quite well. However, on my slightly outdated but hardly obsolete workstations, the results ended up being rather different.

A few years ago, when I migrated my company’s workstations from Windows 2000 to XP, there were problems too, but when I finished working through them, we had better-performing workstations than we'd had before. With Vista, I can’t be so quick to say the same is true. After extensive—and expensive—hardware upgrades, a complex migration process, and the deployment of SP1, the computers are running well now. Are they running noticeably faster than they did with XP? No, but they are no longer noticeably slower either. Is Vista better then XP? It’s been a gradual adjustment process, but I think most of my users would now say yes. For many of them, it’s the small things they have grown to like. They like the refinements to the Windows Explorer interface. They like the Windows Sidebar gadgets. They love the improved search engine.

As an administrator, I’ve grown to appreciate Vista’s way of doing things. I can appreciate that XP is reaching the end of its operational life and will eventually be discontinued. With Vista, I have far fewer updates to evaluate and deploy. The driver situation has improved considerably, and Vista also has a large number of new Group Policy settings, which I’m using to deploy shared printers to workstations and perform other remote configuration tasks. In addition, I’m using Windows BitLocker Drive Encryption to protect the data on the laptops we provide to outside salespeople.

Would I do it all again if I knew then what I know now? I probably would, but in a far more cautious manner. I’d have to say that my biggest mistake was in not doing enough testing on the actual workstations my end users would be using. I would have waited at least until the SP1 release and until I found drivers for all my hardware. As mentioned earlier, I also would have selected video adapters based more on their WDDM driver support than on their ability to support Aero.

Oh, well. You live and learn.