Does anyone upgrade anymore?

It doesn't necessarily speak ill of the upgrades not being purchased, it's an industry phenomenon and a sign of scrutiny in IT budgets that may have been lacking in years past.

Solaris 9, for instance, is not a bad product--far from it. I think Solaris is just a prominent victim of factors rampant in the industry: People aren't spending like they used to, especially since sales of Solaris historically have tended to come as part of expensive hardware deployments. But it's also a victim of its own technical successes.

In September, my favorite monthly news source, the Netcraft Web server survey, took some new angles on their monthly scan of HTTP servers on the Internet, this time classifying by IP address and looking at operating systems. Some interesting Solaris patterns showed up. There were twice as many Solaris 2 and 7 systems as Solaris 8 systems, and fewer than 1,000 IP addresses with Solaris 9. Windows .Net Server had almost as many addresses.

Sun is hardly alone in this phenomenon. Many users aren't satisfied only with their good products like Solaris 8, but also cling to Windows 98 and NT4. When I suggested that NT4 users would have to upgrade relatively soon, I was bombarded with e-mails and TalkBacks from happy NT4 users who were damned if they were going to leave it. (I must admit I really didn't expect this; from my earliest experiences with Windows 2000 I've disliked NT4.)

Dig into the Netcraft site some more and you'll see the evidence of NT4's surprising staying power. The report on FTSE 100's web servers shows 23 of them running Windows NT4 (or Windows 98, but I doubt any of them are actually running 98). The Nasdaq 100 report shows only seven NT4 sites, although that's still more than I would expect.

It's worth reviewing some of the limitations in the Netcraft survey: It only looks for HTTP servers, so if your Solaris 9 system is on the Internet without an HTTP server it doesn't show up in the survey. It also doesn't find systems that are not exposed to the Internet. And remember that Solaris 9 facilitates a continuing trend towards server consolidation, so it's possible that a larger number of users are being served by a smaller number of Solaris machines.Sun is still upbeat about Solaris 9's progress at this point. I asked them how many people were buying Solaris 9 and they said: "At this point Solaris 9 adoption seems to be vastly outstripping where Solaris 8 was at the same time--by a factor of about three to one according to the registration numbers. We have over 300,000 registered licenses of Solaris 9 under 4 months into its lifecycle--this is nothing short of astonishing. I have asked the folks who look after this information if it's a mistake, and they assure me it is not; Solaris 9 seems to be hotter than any of us had realized. Note that these numbers are all before we have released Solaris 9 for x86." And the October Netcraft Survey shows a large increase in Solaris 9 usage, mostly from one hosting company.

Those are impressive numbers. What explains the discrepancy with the Netcraft numbers? Perhaps, as I said before, Solaris 9 is being used on systems that are not accessible to the survey. Perhaps some of the over 300,000 licenses aren't in actual use yet. The latter point isn't as important in a way.

Upgrade revenue hasn't historically been as major an issue for Sun as it is for Microsoft, but letting people get good and ready to buy new product just won't do the job anymore. Their most important mission now is to find new ways to sell Solaris. The most prominent example of this is the recent announcement that they would unbundle Solaris x86.

The actual product will be cheap: $99 for a single CPU desktop and there will even be a $20 time-bombed version of it. But in Sun tradition, service will be optional but expensive ($75 per month for desktop systems and $1,275 a year for lower-end servers according to reports). If Sparc/Solaris is losing market share to lower-cost x86/Linux systems, these prices won't grab back all the business. But it's still better than Sun trying to compete with Linux on price; as Steve Ballmer says, that's not a battle they're going to win.

And just as Microsoft has product lifecycle policies, so does Sun. These policies first withdraw the product from sale through various channels, and then slowly withdraw support, including bug fixes. Sure, it's partly to keep us buying new versions, but it's also just plain reasonable that software companies should be able to move resources from supporting out-of-date products.

Incidentally, even Red Hat has policies such as this: "Important: Red Hat's errata release policy supports the two most recent major product releases. All minor releases in the current product release cycle are supported as is the final release of the prior product release cycle. Currently supported products are indicated below. If you are using an unsupported product and require current software updates, please upgrade to a supported product." The oldest version they support is Red Hat 6.2, which is what, two years old? Linux users have been more eager to upgrade, perhaps because the actual software is free.

It looks like we're headed for a conflict in the long term between the vendors who want to continue selling new versions and customers who want to continue running old versions that work just fine. The problem is getting worse, ironically, as products get better. In the long term, I think it's dangerous for an enterprise to rely in any significant way on products that aren't supported. For that reason it's a good idea for you to start developing a plan for those Solaris 2 systems. If you're not going to move them on to something else, they'd better be expendable.

If you're running an older OS version, what would convince you to upgrade? TalkBack below or e-mail us with your thoughts.