"H.Oura FCT Headquarters??" wrote:
>
> Hi,
>
> I've calculated data transmission rate of my SCSI-2 disk.
No, you haven't. You've calculated the combined transfer period of
128KB + the overhead and latency of a single command creation,
initiation, response, and teardown. Your numbers would be wildly
different if you say sent 2 commands at once and saved the time until
both came back. With Tagged Command Queueing, you may have one command
complete in 24msecs and find that two simultaneous commands might very
well complete in say 28msecs. In short, the code path from sg.c to disk
and back is *way* too long and complex to let the command completion
time infer *anything* about drive bandwidth capabilities based on the
type of test you performed.
> My SCSI
> hardwares are : Adaptec AHA2940UW + Fujitsu MAB3091(Wide SCSI-2 disk).
> I repeatedly wrote 128KB data to MAB3091 using SG interface, and counted
> the time between write(sending CDB) and read(receiving status) call.
> The average time for 128KB data transmission is 24msec.
> So, the data transmission rate is 5.3MB/sec.
>
> I know some kind of overheads are included, but it's too slow for Wide
> SCSI-2 (peak performance(?) of Wide SCSI-2 is 20MB/sec, I guess).
>
> Any suggestions ?
Nope, I'm late and gotta go catch a plane, so I don't have the time to
explain any better tests. A quick start though might be to issue a
series of TEST_UNIT_READY commands to isolate the command startup,
initiation, completion, and teardown latencies. Get a msec measurement
on that command, then subtract it from 24msecs and that will give you
how much time was spent sending the 128K to the drive. Of course, if
the drive is actually writing to media before signaling completion, then
that may add significantly to the number and therefore skew the results
somewhat. Later.
--
Doug Ledford <[EMAIL PROTECTED]>
Opinions expressed are my own, but
they should be everybody's.
-
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to [EMAIL PROTECTED]
