Since you mentioned problem does not happen on Windows, what do you use
to emulate the workload (iometer ?)
CrystalDiskMark, HD Tune, ATTO Disk Benchmark
Tested for 2-3 hours, but got stable results. Directly afterwards booted
Live CD with Linux and immediately got poor results.
and did you see the same behaviour with other Kernel versions like
2.6.32.
I did not test this, as this kernel version wouldn't work for me anyway.
If you think it would be very important datapoint, I can try.
You may want to check the per device queue configuration, lsscsi -L
"queue_depth=256", same for all disks
--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html