The machines implement a real-time missile simulation and data
acquisition system that gets (literally) forklifted into a semi,
driven cross-country, dragged up a mountain, and is still expected
to work flawlessly. Our latest exercise demonstrated that the
(U320) SCSI drives/subchassis/cabling isn't handling the jiggling
very well. Multiple systems had problems booting almost every day
and the current RAID system developed enough issues that it'll have
to be completely reworked.
Lather, rinse, repeat.
The thinking is that we can approach this on two fronts - fewer
overall disks and fewer physical connections to the remaining disks
(SATA or maybe SAS) in a RAID system.