I was on a conference call discussing the Microsoft product lifecycle with several coworkers and our Microsoft-assigned support engineers when someone asked if a server version of Windows 10 was going to come out.
The Microsoft rep said no comment. Then I chimed in.
“We need to assume they will release a server version, probably about six months after the desktop version, and we need to start testing and preparing to deploy it when it comes out,” I said.
“Shouldn’t we wait for Service Pack 1?”
I went in for the kill.
I don’t know where the Service Pack 1 advice came from. Service Pack 1 for Windows NT 4.0 was a train wreck–less reliable than the beta of NT4 was, and with Windows 2000, nobody I knew bothered to wait. We were doing well to get people to wait for the production release–they’d have deployed the beta if they could have gotten away with it. And it was with Service Pack 2 that Windows XP came into its own.
I told a story from the Windows 2003 days. It had just come out, we had it running in the lab, and it was smooth. It was quicker than 2000 on the same hardware, had a nice uptick in reliability, and I couldn’t find any major reason to dislike it, which was unusual because I was a notorious Microsoft critic in those days. It’s all I was known for. My other colleague who’d spent a lot of time with 2003 liked it too.
I was in a meeting with a couple of directors and the talk turned to Windows 2003. One of the directors said Gartner was saying to wait for Service Pack 1. I shook my head.
“This is Service Pack 1,” I argued.
I got a couple of puzzled looks. “This is Windows XP with a few registry tweaks to make it a server, plus whatever bug fixes they found in between the two release dates,” I said. “And XP is just 2000 with a tweaked user interface and better plug-and-play. This thing’s better than what we’re deploying right now.”
It took some talking, but I managed to persuade them to let us deploy a few 2003 boxes to production. And it didn’t let us down. A few months later one of those directors thanked me.
And 2003, of course, became entrenched. Every shop I’ve been in since had a huge fleet of 2003 boxes. And now that’s a problem seeing as 2003 support is ending soon.
There hasn’t been anything wrong with Server 2008 or 2012 either, except that 2003 was good enough that nobody wanted to switch. So now we’re all falling all over ourselves rather than upgrading at a sustainable, disciplined, measured pace.
My director, senior director and I were talking about this problem on Friday. We all agreed we have to get faster at migrating to the new server versions.
My plan is the most aggressive one of the bunch. We can count on Microsoft giving us at least 10 years, and a new version coming out in about five. So I’m convinced the best thing to do is spend five years ramping up the new version and ramping down the previous one. About 2 1/2 years in, we would have about the same number of new boxes and previous-version boxes, and five years in, we’d be shutting down the last of the previous-version boxes. And then we would start the cycle over again.
Too aggressive for you? Don’t talk to my former boss then, who’s now the CTO at a small credit union in Illinois. His policy is to rebuild every Windows server every two years, unless it gives you trouble, in which case rebuild it after one year. And when you’re rebuilding that often, he’s going to slip an OS upgrade in there one of those years.
At least I’m proposing you keep your server five years.