Over the years, my perspective on computer upgrades has changed significantly. There was a time back in the mid-to-late 90s when I would go out and obtain a new motherboard and processor every 15-18 months, and swap out the components on one of my systems.
In the early going, this wasn’t too bad, as the performance improvement was substantial enough to offset the scraped knuckles (remember those old computer cases?) and the inevitable troubleshooting effort that followed motherboard replacements. This left me with some spare parts (optimistically)that I could use as backups for my other systems, or perhaps sell/give to someone else who still could benefit from them.
Later, my brother and I would go to the various computer shows and get barebones systems, which lowered the overall migration risk, because then we only had to worry about migrating over our hard drives, and maybe the video card. Still, these approaches to upgrades left us with “spare” parts that sometimes never got used. (I still have a ton of parts that don’t fit into anything that anyone would care to own at this time).
Essentially, we’d end up with 1.2 systems for every system upgraded. I knew a lot of people that would sell off any parts they had remaining, but that was most effective with an upgrade cycle of 6-9 months, where you’re always buying the latest and the greatest… (Don’t get me started on how pointless I think it is to be upgrading every 5 days before you can even experience any benefit from the last upgrade).
While we’re sort of on the topic of trends I never really subscribed to, I have to say that on a whole, I was always more focused on overall network performance and interoperability than on individual machine performance. This means, that except for occasional benchmarking, I was never really into the whole “I got 2.1% more UltraFoozleMarks with my latest overclock!!” craze that was so prevalent in late 90s through about 2004. And while I did my fair share of overclocking, it was more a means to an end, than the end itself. At one point it was so bad that many folks were spending more on cooling than they would have if they had simply purchased the next higher processor.
Anyhoo, let’s get back to the subject at hand. It started to become apparent to me that one of the biggest drawbacks to in-place upgrades was time. You basically had to complete the upgrade as fast as possible, because while it was ongoing, you didn’t have use of your system. And you hoped that there were no complications when you were finished, because then you’d be forced to downgrade again to be sure all the components were working properly.
(I don’t even want to think about how many times an upgrade failed, and then the backout to the old hardware failed as well!)
So, what was the alternative? In 1998, I decided to upgrade one of my servers, and rather than perform an in-place upgrade to save a little money, I went to a computer show and purchased a new case, memory, hard drives and video card. I added a NIC I had salvaged from some previous upgrade, and I started to build my new server. It was refreshing to be able to take my time with the upgrade, because I wasn’t touching the existing system. No undue pressure to finish the upgrade in one night or weekend to minimize downtime.
Once I had all the data copied over that I needed, I was able to set aside some time one evening to perform the cutover (it was a new domain controller, for a new domain). After a week or so of using the new system, I repurposed the old one – again without suffering much downtime. And so, I began to use this method whenever I needed to upgrade. My last 7 or 8 upgrades have all followed the same path, and I usually take two or three weeks to fully migrate over to using the new system, whether it’s a desktop or server. (The only exception to this was the recent desktop for my kids, who applied a substantial amount of daily “when is it going to be finished” pressure to ensure that I delivered their machine to them in under a week).
After a swap-grade, you have the option of repurposing the old system for some new functions, keeping it as a backup to the new system, or selling/donating it. I’ve used all of these alternatives at some point in the past 8 years, and the result has been far less involved (or fraught with peril) than the in-place upgrade. In the end, I see lots more benefit to ending up with 2 systems, even if one of them is a little aged, than ending up with 1.2 systems.
And now that my family has turned into a set of demanding users of technology (just try to take down a server in the middle of the day for maintenance at my house), swap-grades are proving to be the only viable approach for me. Oh, I’ll still add memory or hard drives to an existing system, or even change out the video card every now and then, but the days of wholesale motherboard changes are gone, gone, gone…
Obviously, this won’t be the best course of action for everyone, but it certainly continues to work for me and my network, and I have no problems recommending that others consider it for their own environments.