Hi,
We depend on I/O scheduler in dm-mpath layer, and underlying
I/O scheduler is bypassed basically.
I/O scheduler depends on queue busy condition to
trigger I/O merge, unfortunatley inside dm-mpath,
the underlying queue busy feedback is not accurate
enough, and we just allocate one request and dispatcch
it out to underlying queue, no matter if that queue
is busy or not. Then I/O merge is hard to trigger.
This patchset sets underlying queue's nr_request as
the queue's queue depth, so that queue busy is figured
out by checking if request is allocated successfully.
>From test result on mq-deadline, sequential I/O performance
is improved a lot, see test result in patch 5's commit log.
Any comments are welcome!
Thanks,
Ming
Ming Lei (5):
block: don't call blk_mq_delay_run_hw_queue() in case of
BLK_STS_RESOURCE
dm-mpath: return DM_MAPIO_REQUEUE in case of rq allocation failure
dm-mpath: remove annoying message of 'blk_get_request() returned -11'
block: export blk_update_nr_requests
dm-mpath: improve I/O schedule
block/blk-core.c | 4 +++-
block/blk-sysfs.c | 5 +----
block/blk.h | 2 --
drivers/md/dm-mpath.c | 30 +++++++++++++++++++++++++++---
drivers/md/dm-rq.c | 1 -
drivers/nvme/host/fc.c | 3 ---
drivers/scsi/scsi_lib.c | 4 ----
include/linux/blkdev.h | 1 +
8 files changed, 32 insertions(+), 18 deletions(-)
--
2.9.5
