Bah!/Shiney
Well, the plan to go with the older motherboard was thwarted by the fact that I don’t actually have a keyboard with a full DIN connector (rather than the more standard PS2/mini-DIN connectors now commonplace). D’oh! So its back to the same old motherboard, but sans the disk that died on me. So, still random stability problems, but now the shoddy disk is gone, at worst the machine will need rebooted to get access to things. Still sounds like a hoover though – I should probably retrieve my old P2 Celeron from the person I loaned it to and use that instead. Still, not too urgent.
I’ve decided to up my backup frequency though – restoring from an older subversion database is a pain, as all the changes made have to be isolated and re-commited to the DB seperately. Way more time consuming than it should be.
On an unrelated note – I liked the case study of EVE’s most recent hardware upgrade: detailed here. Sounds like a case of ‘can we have this new shiney bit of kit please?’ followed by ‘OMG thats shiney – everything works great now!’. Of course, the fact that you have to upgrade to faster hardware to sustain more users points to a flaw in their scalable design. In theory, a truly scalable design means that you can support more users by simply increasing the number of processing units of the same type available; otherwise you’ll eventually run into a hardware wall (disk drives can only get so fast). Still, I’m sure it won’t be an issue for a long while – increasing the power of the bottlenecked process by a factor of 10 in theory means allowing 10 times as many users before it becomes an issue again, long enough for another item to become the bottleneck first…