> If its an IO problem the first and easiest thing to do is (probably) look at
> your disk subsystem. You can easily achieve higher disk IO by increasing the
> number of disks and implementing something like RAID1+0.
Or you can be logical about it and try to determine whether the IO
performance is a symptom or a cause. If there are queries that don't
have good indexes, "add correct indexes" is a smarter solution than
"add disks." Indeed, even the IO usage can be a red herring.
I suggest a more systematic approach to the problem, such as Method R.