* [PATCH] block/023: performance test on queue creation & cleanup
@ 2018-07-04 5:29 Ming Lei
2018-07-25 21:33 ` Omar Sandoval
0 siblings, 1 reply; 3+ messages in thread
From: Ming Lei @ 2018-07-04 5:29 UTC (permalink / raw)
To: Omar Sandoval; +Cc: Jens Axboe, linux-block, Ming Lei
SCSI may have lots of channels, targets or LUNs, so it may
take long time for creating and cleaning up queues.
So introduce block/023 and uses null_blk to run this test
on both blk-mq and legacy mode, then compare both and check
the difference.
Signed-off-by: Ming Lei <ming.lei@redhat.com>
---
tests/block/023 | 101 ++++++++++++++++++++++++++++++++++++++++++++++++++++
tests/block/023.out | 2 ++
2 files changed, 103 insertions(+)
create mode 100755 tests/block/023
create mode 100755 tests/block/023.out
diff --git a/tests/block/023 b/tests/block/023
new file mode 100755
index 000000000000..5a8f323bd42a
--- /dev/null
+++ b/tests/block/023
@@ -0,0 +1,101 @@
+#!/bin/bash
+#
+# Performance regression test on blk-mq queue creation & cleanup. Measure
+# the time for creating & cleaning up null_blk queues in both legacy and
+# blk-mq mode, then compare both & check if there is performance regression.
+#
+# Copyright (C) 2018 Ming Lei
+#
+# This program is free software: you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation, either version 3 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program. If not, see <http://www.gnu.org/licenses/>.
+
+. tests/block/rc
+
+DESCRIPTION="performance regression test on null_blk queue creation & cleanup"
+QUICK=1
+
+requires() {
+ _have_module null_blk
+}
+
+remove_null_blk_mod() {
+ delay=$1
+ while true; do
+ modprobe -r null_blk > /dev/null 2>&1
+ [ $? -eq 0 ] && break
+ sleep $delay
+ done
+}
+
+test_null_blk_queue_perf() {
+ QUEUE_MODE=$1
+
+ remove_null_blk_mod 1
+
+ start_time=`date +'%s'`
+ modprobe null_blk nr_devices=4096 queue_mode=$QUEUE_MODE
+ end_time=`date +'%s'`
+ create_time=`expr $end_time - $start_time + 1`
+
+ start_time=`date +'%s'`
+ remove_null_blk_mod 0.1
+ end_time=`date +'%s'`
+ cleanup_time=`expr $end_time - $start_time + 1`
+
+ echo $create_time $cleanup_time
+}
+
+check_diff() {
+ src=$1
+ dst=$2
+ msg=$3
+
+ diff=`expr $dst - $src`
+ [ $diff -le 10 ] && return
+
+ result=`echo "scale=0;($diff - $src * 0.4) / 1" | bc`
+ [ $result -ge 0 ] && echo $msg
+}
+
+check_limit() {
+ src=$1
+ dst=$2
+ msg=$3
+
+ result=`echo "scale=0;$dst - $src * 20" | bc`
+ [ $result -ge 0 ] && echo $msg
+}
+
+test() {
+ echo "Running ${TEST_NAME}"
+
+ rq_time=($(test_null_blk_queue_perf 1))
+ mq_time=($(test_null_blk_queue_perf 2))
+
+ echo ${rq_time[0]} ${mq_time[0]} >> "$FULL"
+ echo ${rq_time[1]} ${mq_time[1]} >> "$FULL"
+
+ msg="blk_mq takes too long(${mq_time[0]}s) to create queues compared with block rq(${rq_time[0]}s)"
+ check_diff ${rq_time[0]} ${mq_time[0]} "$msg"
+
+ msg="blk_mq takes too long(${mq_time[1]}s) to cleanup queues compared with block rq(${rq_time[1]}s)"
+ check_diff ${rq_time[1]} ${mq_time[1]} "$msg"
+
+ msg="block rq takes too long(${rq_time[1]}s) to cleanup queues compared with creating queues(${rq_time[0]}s)"
+ check_limit ${rq_time[0]} ${rq_time[1]} "$msg"
+
+ msg="blk_mq takes too long(${mq_time[1]}s) to cleanup queues compared with creating queues(${mq_time[0]}s)"
+ check_limit ${mq_time[0]} ${mq_time[1]} "$msg"
+
+ echo "Test complete"
+}
diff --git a/tests/block/023.out b/tests/block/023.out
new file mode 100755
index 000000000000..4a11db1f10fe
--- /dev/null
+++ b/tests/block/023.out
@@ -0,0 +1,2 @@
+Running block/023
+Test complete
--
2.9.5
^ permalink raw reply related [flat|nested] 3+ messages in thread
* Re: [PATCH] block/023: performance test on queue creation & cleanup
2018-07-04 5:29 [PATCH] block/023: performance test on queue creation & cleanup Ming Lei
@ 2018-07-25 21:33 ` Omar Sandoval
2018-07-29 22:43 ` Ming Lei
0 siblings, 1 reply; 3+ messages in thread
From: Omar Sandoval @ 2018-07-25 21:33 UTC (permalink / raw)
To: Ming Lei; +Cc: Omar Sandoval, Jens Axboe, linux-block
On Wed, Jul 04, 2018 at 01:29:56PM +0800, Ming Lei wrote:
> SCSI may have lots of channels, targets or LUNs, so it may
> take long time for creating and cleaning up queues.
>
> So introduce block/023 and uses null_blk to run this test
> on both blk-mq and legacy mode, then compare both and check
> the difference.
>
> Signed-off-by: Ming Lei <ming.lei@redhat.com>
> ---
> tests/block/023 | 101 ++++++++++++++++++++++++++++++++++++++++++++++++++++
> tests/block/023.out | 2 ++
> 2 files changed, 103 insertions(+)
> create mode 100755 tests/block/023
> create mode 100755 tests/block/023.out
Hi, Ming, is this a regression test for "blk-mq: remove
synchronize_rcu() from blk_mq_del_queue_tag_set()" and "blk-mq: avoid to
synchronize rcu inside blk_cleanup_queue()"?
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH] block/023: performance test on queue creation & cleanup
2018-07-25 21:33 ` Omar Sandoval
@ 2018-07-29 22:43 ` Ming Lei
0 siblings, 0 replies; 3+ messages in thread
From: Ming Lei @ 2018-07-29 22:43 UTC (permalink / raw)
To: Omar Sandoval; +Cc: Omar Sandoval, Jens Axboe, linux-block
On Wed, Jul 25, 2018 at 02:33:14PM -0700, Omar Sandoval wrote:
> On Wed, Jul 04, 2018 at 01:29:56PM +0800, Ming Lei wrote:
> > SCSI may have lots of channels, targets or LUNs, so it may
> > take long time for creating and cleaning up queues.
> >
> > So introduce block/023 and uses null_blk to run this test
> > on both blk-mq and legacy mode, then compare both and check
> > the difference.
> >
> > Signed-off-by: Ming Lei <ming.lei@redhat.com>
> > ---
> > tests/block/023 | 101 ++++++++++++++++++++++++++++++++++++++++++++++++++++
> > tests/block/023.out | 2 ++
> > 2 files changed, 103 insertions(+)
> > create mode 100755 tests/block/023
> > create mode 100755 tests/block/023.out
>
> Hi, Ming, is this a regression test for "blk-mq: remove
> synchronize_rcu() from blk_mq_del_queue_tag_set()" and "blk-mq: avoid to
> synchronize rcu inside blk_cleanup_queue()"?
Yeah, it is.
It also shows that destroying queue may take much more time than
creating queue, and seems no solution for this one now.
Thanks,
Ming
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2018-07-29 22:43 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2018-07-04 5:29 [PATCH] block/023: performance test on queue creation & cleanup Ming Lei
2018-07-25 21:33 ` Omar Sandoval
2018-07-29 22:43 ` Ming Lei
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.