Doug Ledford <[EMAIL PROTECTED]> wrote:
>> The only question I have is whether or not there is a circumstance where
the
>> queue depth is more logically considered to be a bus concept rather than
a
>> device concept.
>
>Not really, that's why we spec the overall controller queue depth during
host
>registration time. That should be sufficient for the bus queue depth, and
>this should be used for individual device queue depths.
I understand device queue depth, but I'm not fully conversant with
"controller
queue depth" and "bus queue depth" so the statement "this should be used
for
individual device queue depths" tweaks my interest.
I've been testing Linux attached to BIG disk subsystems and found that
Linux
(w/ aic7xxx defaulting to no TCQ) performance did not match that seen by
other non-Linux hosts. The performance gurus for the subsystem recommended
device
TCQ values of "70-80"..... which fixed the performance problem.
Please keep this type of variety in mind when deciding what values are
sufficient, or continue to provide ways to adjust the values in customer
production environments.
Thanks,
Steve
-
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to [EMAIL PROTECTED]
