The Upgrade Saga

One of the things I like about using Conversant, aside from the fact that it’s the only content management/groupware system that can meet my personal knowledge management needs, is that Seth Dillingham always goes above and beyond what most people would do (or at least what most people would do from my experience).

A few weeks ago I ran out of disk space on the 18 gig hard drive that hosted this site. With more than 70,000 messages in the database and with my sites serving up 25,000+ page views daily, I really needed a new hard drive.

But Seth suggested that other upgrades would also improve performance, and since hardware is all but dirt cheap these days I decided to overhaul this server. So out would go the 1 ghz Athlon, the 768 mb of RAM, and the 18 gig hard drive, and in would come a 2.4 ghz P4, 1 gig of RAM, and a 60gb 7200 RPM Barracuda drive. No problem, right?

Except that pretty much every time Seth and I have done anything with the hardware something goes wrong (typically bad parts).

So I order the components and have them shipped to Seth, and he’s busy putting it together. Except the old server had an ISA video card and the Gigabyte motherboard I’ve ordered doesn’t have any ISA slots, so Seth runs out and picks up a cheap used one (and you have to keep in mind here that Seth lives in the middle of nowhere).

And then comes gotcha #2 — although it isn’t explicitly documented anywhere that I could find, the board is shipped such that it has to be booted at least initially with an AGP 4X card. Yet another trip.

So Seth’s got the AGP 4X card, but when he boots the system the hard drive fails. It looks like maybe there’s something wrong with the cable or maybe a jumper issue on the board. Nope — it turns out to be a power supply issue. We were going to use a 250W power supply which is simply not putting out enough juice.

This means Seth makes a 90-mile round trip drive to CompUSA on Sunday to pick up a nice 400W power supply (and, of course, these additional parts are adding to the expense, however moderately). On that trip, the transmission in Seth’s truck starts acting funny. As I told my wife, “I think I just broke Seth’s truck.”

So, the server is finally working and Seth’s installing and copying all of the software and data, but will his truck make the 90-mile or so trip to the ISP? Fortunately, he takes the car to his mechanic and they find nothing wrong with it.

But now we’re long past the time to make the drive, which is why the server doesn’t get back up until this afternoon.

Whew!

The end result, however, was worth it, though. Most pages render in under a second on my broadband connection which is a pretty amazing feat given how much of the content is generated dynamically. Some of the pages that rely on a lot of intensive search of a large message base went from taking 15 to 20 seconds to render to down to 3 or 4 seconds. Since the sites I run are first and foremost for my own reference purposes this is a huge time saver and increases the value of the site to me above and beyond the improved user experience for everyone else.

The only problem? Human beings are too good at adapting to such changes in speed. Give me a couple weeks and I’ll be pricing 15K SCSI drives thinking I might get that 3-4 second load time for the resource-intensive pages down to 1-2 seconds (and then, imagine taking two of them and putting them in a RAID array — of course, I’m not sure if Seth’s truck could take that sort of upgrade!!) And of course it won’t be too long before we see 4 ghz processors . . .

Post Revisions:

There are no revisions for this post.

Leave a Reply