The Quantum drives are known to have problems in how they handle tagged
queueing commands (various firmwares on various drives have different
problems, most related to the issue of releasing tagged commands as complete
when they are still occupying resources on the drives controller). In any
case, I usually don't recommend over 14 commands per lun on Quantum drives
of any variety (where as my Seagate Barracuda drives are happy with 32 or 40
if I want). I would try reducing the commands per lun value and check your
results again.
The most recent L915 firmware for the Quantum Atlas drives corrects all
problems with the tagged queuing implementation that I've encountered or heard
of. The XP34300W with 81HB ran fine for me with *no* timeouts in 18 months at
a queue depth of 28. The L912 firmware is definitely buggy, and should be
replaced with L915 with is available from Quantum's FTP site.
The Quantum Atlas II LXY4 firmware also has a bug with tagged queuing that
causes the drive to become catatonic in some cases, leaving uncompleted
commands waiting indefinitely until a command is aborted and a new command is
sent. I reproduced this problem on a SCSI analyzer at BusLogic in June,
reported it to Quantum when we confirmed it was the disk drive at fault, and
then demonstrated it at Quantum's facility a week later. Quantum has fixed the
bug, but no version of firmware is yet available with the fix.
As with the unreleased versions of firmware for the Quantum Tempest and Stratus
drives I announced recently, I will seek to make the updated Atlas II firmware
available from my web page as soon as it is available.
Leonard