On 12/01/2021 07:23, Hannes Reinecke wrote:
So it seems to me that the block layer is merging more bios per
request, as averge sg count per request goes up from 1 - > upto 6 or
more. As I see, when queue depth lowers the only thing that is really
changing is that we fail more often in getting the budget in
scsi_mq_get_budget()->scsi_dev_queue_ready().
So initial sdev queue depth comes from cmd_per_lun by default or
manually setting in the driver via scsi_change_queue_depth(). It seems
to me that some drivers are not setting this optimally, as above.
Thoughts on guidance for setting sdev queue depth? Could blk-mq
changed this behavior?
First of all: are these 'real' SAS SSDs?
The peak at 32 seems very ATA-ish, and I wouldn't put it past the LSI
folks to optimize for that case :-)
Can you get a more detailed picture by changing the queue depth more
finegrained?
(Will get you nicer graphs to boot :-)
They're HUSMM1640ASS204 - not the fastest you can get today, but still
decent.
I'll see about fine-grained IOPs vs depth results ...
Cheers,
John
