One day, I came back to my desk after lunch and my boss cornered me. “You know about these things. What’s Half-Life?”

I hesitated for a minute. “I’m pretty sure it’s a computer game.”

It wasn’t a game I’d played. If it isn’t Railroad Tycoon, I’m probably not interested. I did some quick digging, confirmed my hazy never-cared-before-then recollection, and got back to him with the verdict: It’s a multiplayer, network-capable first-person shooter game. Nothing at all like Railroad Tycoon. So I wasn’t interested. Well, not in the game. But I wanted to know why he was asking. He didn’t seem like the type who would be interested either.

That network-capable part turned out to be key.

As it turned out, there was a small department with a handful of users. It might have been four, it might have been a couple dozen. They had their own T1 line, because they didn’t think the shared T1 line everyone else used would provide them with the throughput they needed.

After a few months, their T1 line proved inadequate and they started complaining that they wanted something bigger. In 1999, there was no legitimate reason for a couple of dozen users to fill up the capacity of a T1 line. Back then there were still entire college campuses that just had a T1 for everything. I remember being surprised when I started work at a shop that put a mere 1,000 people on a T1 connection.

Today, your typical $30 Internet connection theoretically delivers T1 speed, if not better. But back then, a T1 cost around $1,500 a month, and if a T1 wasn’t adequate, you had to step up to a T3, and you were looking $26,000-$32,000 per month for that.

So a few heavy hitters came in to investigate before anyone went and spent the money to upgrade to a T3 connection. Wise move.

The department had its own server, and their own crackerjack system administrator to set it up and take care of it. I don’t know if the investigation started there, but it sure ended there. On the server, they found a process containing the words “half life” using a small amount of CPU power.

Not being something one normally found on something running Windows NT Server 4.0 at work, it obviously attracted some attention. Of course, the local system administrator wasn’t buying it. The problem couldn’t be anything he did, and certainly not something using 10% or so of the CPU.

“Look!” the local administrator said. “That’s not using much power. Look at that! 87% of the CPU is going to ‘System Idle Process!’ That must be it!”

“I need more CPU power for my System Idle Process” became a running joke for years after that.

But sure enough, after shutting down the Half-Life process, which turned out to be a Half-Life server–and an extremely popular one at that–the usage on the T1 dropped to a level that was acceptable to the users in that department.

At that point, it became an HR matter. Like my boss said, taking a pen here, a paper clip there is one thing. Taking a T1 is another. Especially if it happens to be a fractional T1, where you’re billed on usage. This one was.

This story is one reason I don’t think it’s a very good idea for departments to run their own servers. IT departments usually don’t make the most popular decisions, but there’s something to be said for experience and accountability.