Windows Migrations Made Simpler

Microsoft’s Windows Server “Longhorn” operating system will be entering the world shortly. And that means Windows Server 2003 is almost out of date, Windows Server 2000 is almost obsolete, and Windows NT Server is little more than a historical curiosity.

If you’re among the 35 percent of organizations still running a Windows NT server, it’s time to think about making a change. When does migration to a newer Microsoft operating system make sense, and what questions must be asked for a successful migration?

That may be Microsoft’s idea of the state of play on the operating system world, but the reality — for those responsible for delivering and managing IT infrastructures — is rather different. If NT is dead, no one seems to have told them. A recent Gartner survey found that 35 percent of organizations still have at least one NT server running somewhere, doing something, according to principal research analyst Jonathan Hardcastle. The total number of NT servers in use today must therefore be quite large: Gartner estimates that of the 20 million Windows servers around the world, about 1 million are NT servers. An appreciably larger number are Windows 2000 servers.

Why are there so many NT servers out there, when the support life cycle for Windows NT Server finished at the end of 2004, and only custom support is available — at a price that doesn’t make sense for a company with a single NT server tucked away in the proverbial closet somewhere? Even Windows Server 2000 is coming to the end of its support life cycle. It is in the Extended Support phase, which essentially means Microsoft continues to provide security hotfixes and paid support but no longer provides complimentary support options, design change requests, and nonsecurity hotfixes.

The obvious answer is the old “if it ain’t broke, don’t fix it” argument, but this makes less and less sense in an era where security is vital.

“If you are not on custom support, it is pretty reckless to be using NT, but you may have some applications which are not exposed on the Web or to outside sources,” says Hardcastle.

This opinion is echoed by Gordon Haff, principal IT advisor at New Hampshire based research house Illuminata. “There was a time when old servers ran forever in a closet with whatever OS and apps they were running, but today, in most circumstances, this is simply not secure. In the prenetworked age it was no big deal, but now it is hard to say that a server is isolated. It is risky.”

What’s true for NT will be true for 2000 in a few years, so organizations with an NT server — unless it really is running in complete isolation — or a 2000 server should begin exploring migration options now.

And “migrate,” as opposed to upgrade, is indeed the correct word in this context. In the case of Windows NT, even moving to another Windows platform is a significant change, not least because of the introduction of Active Directory. Moving a Windows Server 2000 server to 2003 or beyond is a less significant change, but the difficulties that may be encountered should not be underestimated.

An interim measure some organizations are taking is virtualizing their NT servers using software from VMware. This has a number of benefits. Moving the server to newer physical hardware increases reliability, and getting NT-based applications to run in virtual hardware takes physical hardware out of the equation when migrating the applications to another operating system running in an identical virtual machine.

Of course, the big question is this: Once you decide to migrate from NT or Windows Server 2000 to another platform, which platform do you chose? On paper, abandoning NT presents the ideal opportunity to consider non-Microsoft alternatives, which in most cases will be one of the enterprise-class distributions of Linux, such as Red Hat or SUSE. But this is not necessarily the case.

“In the past we thought NT4 users might be candidates to move to Linux, as maybe they didn’t want the complexity of later versions of Windows with Active Directory. In reality, we haven’t seen that,” says Gartner’s Jonathan Hardcastle.

There are probably two main reasons for this. First, it’s all about apps. If you have custom-engineered VB6 apps running on an NT server, it may not be the easiest task to get them to run on a newer Windows Server product (or rewrite them in the .Net framework), but you can bet it will be harder to port them to Linux. And second, if you’re a Microsoft-only shop, switching even relatively simple tasks like file and print services to Linux is far from trivial.

“Moving to a mixed Unix/Linux environment is hard, as you won’t have the necessary Linux skills, and integration of the two OSes is a fairly significant move,” says Haff. “One of the characteristics of Microsoft is that it tends to monolithics. It’s better than before but, it still doesn’t play that nicely with other operating systems.”

If, on the other hand, you already have a mixed environment of Windows and Linux or Unix, then getting rid of your last NT server doesn’t automatically mean migrating to another Windows platform. An NT-based file and print server, for example, could be reasonably replaced by a Samba server and still serve Windows clients.

The final question worth asking is this: If you are sticking with Microsoft, do you migrate to Windows Server 2003, which is reasonably mature and still in the initial Mainstream Support phase, or wait until later this year for Server “Longhorn”?

“I think it probably makes more sense to look at the stable Windows Server 2003 rather than wait for Longhorn,” says Hardcastle. “To be prudent, you need to wait for the first Longhorn service pack which is perhaps quite some time ahead, while Windows Server 2003 is a reasonably well-proven release.”