On 8/26/21 5:05 PM, Jens Axboe wrote:

quoted

Here is an overview of the tests I ran so far, all on the same test
setup:
* No I/O scheduler:               about 5630 K IOPS.
* Kernel v5.11 + mq-deadline:     about 1100 K IOPS.
* block-for-next + mq-deadline:   about  760 K IOPS.
* block-for-next with improved mq-deadline performance: about 970 K IOPS.

So we're still off by about 12%, I don't think that is good enough.
That's assuming that v5.11 + mq-deadline is the same as for-next with
the mq-deadline change reverted? Because that would be the key number to
compare it with.

A quick attempt to eliminate the ktime_get_ns() call from
__dd_dispatch_request() improved performance a few percent but not as much
as I was hoping. I need a few days of time to run these measurements, to
optimize performance further and to rerun all functional and performance
tests.

Thanks,

Bart.