February 07, 2007

Our server

At some point around 2:00 a.m. this morning, Language Log's elderly linux server began to experience some disk problems, which caused apache to fail. I went in this morning and "fixed" it (at the cost of a distressingly large number of fsck repairs, whose exact nature and consequences I don't know -- it's been a while since I wrangled inodes, and I don't have time to check the details anyhow).

As of about 9:00 a.m., the machine was up and running again, and the relevant bits seem to be OK. But I'll take this as a sign that I should make some more stable hosting arrangements. Meanwhile, let's keep our fingers crossed. If we go off the air again, you'll know why.

And while we're on the subject, let me tell you the story of this venerable machine.

It's a Dell Optiplex GX-240, which was purchased almost exactly 5 years ago, in early 2002, as part of a set of eight or ten machines that we bought for the public areas at the Institute for Research in Cognitive Science at Penn, where I was then director. We bought the machines (for about $1,000 each, as I recall) to replace a variety of really antique computers (some superannuated Suns, for example). We bought inexpensive, compact-format desktop machines from Dell -- as pictured above -- with Windows 2000 installed.

But as a result of reconfiguring our space to make more offices, some of the "public terminal stations" (as they used to be called back in the neolithic era) were eliminated. And as result, a couple of those Dells were not needed.

At about the same time, we started a new research project on biomedical information extraction. It was one of a pair of projects at Penn funded by NSF under the Information Technology Research (ITR) program. One of them was on language modeling of macromolecules, and the other was on information extraction from biomedical text, and so those of us who were involved in both called them ITR-L and ITR-E for short.

The ITR-E project needed a web server, for trying out new kinds of medline searches and the like. We took steps to order and set up a new machine, but there were various delays in the process. So, spying the two unused Dell boxes stacked in the corner, I commandeered one of them, and we plugged it in, installed linux, named it "itre" (after our ITR-E project), and used it until better arrangements were made.

Now fast forward to July of 2003. Geoff Pullum and I decided to give this blogging thing a try. I considered various options. The free weblog sites then available seemed too restrictive to me, and somewhat unreliable (though I might well have been wrong about this). I talked with some sysadmins of existing web servers at Penn, who were politely unenthusiastic about adding to their duties on some faculty member's whim. Paid hosting would cost money, and would require research to select a site, and negotiation to set it up, and all of that seemed like too much trouble for an experiment that we'd probably abandon after a couple of months.

Once again, my eye fell on that old Optiplex GX-240. It was idle again, sitting on an unused desk in a group office at IRCS, since by then the bioie project had found another digital home. This looked like the quickest and easiest path -- and being an inveterate taker of easy paths, I took it. I downloaded and installed Movable Type, picked the first default layout that didn't look unacceptably lame, and tried a few test posts. It worked!

Geoff Pullum, as he had agreed to do, started posting a couple of months later. In October of 2003, we started to get somewhat serious about the enterprise, recruiting Geoff Nunberg and then a series of other colleagues to join us. (Quite a few said "no", or agreed to participate and then didn't, but we've managed to pick up a congenial crew who like the format and have interesting things to say in it.)

In any case, Language Log's server has been running, essentially untouched, since it was set up in the summer of 2003. We've gone off the air a handful of times -- a couple of power outages, a couple of network outages -- but nothing more than a reboot has ever been needed. Before now.

I guess it's time to do something different. I'm not sure just what, but it's time for a change.

Posted by Mark Liberman at February 7, 2007 11:34 AM