Hmmm...one crazy setup I could think of:
Two 6 foot veritcal racks, custom designed to hold a 760MP mobo with dual 1.5GHz Palominos, one per every 2u of rack space. A third rack between the two, holding custom-designed 2u RAID chassis. Each RAID chassis would hold three hot-swap SCA hard drives and two hot-swap controller modules, each controller module having a SCSI connection to its host for data transfer and a serial connection to deliver heartbeat/failover signals. Each RAID chassis would have each controller module connected to a system in one of the other two racks, so that every RAID chassis was connected to the corresponding 2u hosts in the other two racks. Each 2u host system would be connected to the 2u host system in the other rack via a serial cable, once again for heartbeat/failover signaling. All hosts would be wired to a single copper Gigabit switch.
When any host system (or its corresponding controller module in a RAID chassis) would fail, the corresponding host system from the other rack (the one connected to the RAID chassis's second controller module) would take over. The hosts would, of course, be clustered, running Linux 2.4.4 with the latest MOSIX patches.
Not sure what I'd use it for, but it would be fun to play with!
I'd still have to keep my <A HREF="http://www.anandtech.com/mysystemrig.html?rigid=5092" target="_new">current rig</A> around though...
Kelledin
bash-2.04$ kill -9 1
init: Just what do you think you're doing, Dave?