Message-ID: <ab662add-6880-ffdb-98cc-7a947374489d@huawei.com>
Date:   Wed, 28 Sep 2022 09:07:46 +0800
From:   Yu Kuai <yukuai3@...wei.com>
To:     Paolo Valente <paolo.valente@...aro.org>,
        Yu Kuai <yukuai1@...weicloud.com>
CC:     Tejun Heo <tj@...nel.org>, <axboe@...nel.dk>,
        Jan Kara <jack@...e.cz>, <cgroups@...r.kernel.org>,
        <linux-block@...r.kernel.org>, <linux-kernel@...r.kernel.org>,
        <yi.zhang@...wei.com>, "yukuai (C)" <yukuai3@...wei.com>
Subject: Re: [patch v11 0/6] support concurrent sync io for bfq on a specail
 occasion

Hi, Paolo

在 2022/09/28 0:38, Paolo Valente 写道:
> 
> 
>> Il giorno 16 set 2022, alle ore 09:19, Yu Kuai <yukuai1@...weicloud.com> ha scritto:
>>
>> From: Yu Kuai <yukuai3@...wei.com>
>>
>> Changes in v11:
>> - keep the comments in bfq_weights_tree_remove() and move it to the
>> caller where bfqq can be freed.
>> - add two followed up cleanup patches.
>>
>> Changes in v10:
>> - Add reviewed-tag for patch 2
>>
>> Changes in v9:
>> - also update how many bfqqs have pending_reqs bfq_bfqq_move().
>> - fix one language in patch 4
>> - Add reviewed-tag for patch 1,3,4
>>
>> Changes in v8:
>> - Instead of using whether bfqq is busy, using whether bfqq has pending
>> requests. As Paolo pointed out the former way is problematic.
>>
>> Changes in v7:
>> - fix mismatch bfq_inc/del_busy_queues() and bfqq_add/del_bfqq_busy(),
>> also retest this patchset on v5.18 to make sure functionality is
>> correct.
>> - move the updating of 'bfqd->busy_queues' into new apis
>>
>> Changes in v6:
>> - add reviewed-by tag for patch 1
>>
>> Changes in v5:
>> - rename bfq_add_busy_queues() to bfq_inc_busy_queues() in patch 1
>> - fix wrong definition in patch 1
>> - fix spelling mistake in patch 2: leaset -> least
>> - update comments in patch 3
>> - add reviewed-by tag in patch 2,3
>>
>> Changes in v4:
>> - split bfq_update_busy_queues() to bfq_add/dec_busy_queues(),
>>    suggested by Jan Kara.
>> - remove unused 'in_groups_with_pending_reqs',
>>
>> Changes in v3:
>> - remove the cleanup patch that is irrelevant now(I'll post it
>>    separately).
>> - instead of hacking wr queues and using weights tree insertion/removal,
>>    using bfq_add/del_bfqq_busy() to count the number of groups
>>    (suggested by Jan Kara).
>>
>> Changes in v2:
>> - Use a different approch to count root group, which is much simple.
>>
>> Currently, bfq can't handle sync io concurrently as long as they
>> are not issued from root group. This is because
>> 'bfqd->num_groups_with_pending_reqs > 0' is always true in
>> bfq_asymmetric_scenario().
>>
>> The way that bfqg is counted into 'num_groups_with_pending_reqs':
>>
>> Before this patchset:
>> 1) root group will never be counted.
>> 2) Count if bfqg or it's child bfqgs have pending requests.
>> 3) Don't count if bfqg and it's child bfqgs complete all the requests.
>>
>> After this patchset:
>> 1) root group is counted.
>> 2) Count if bfqg has pending requests.
>> 3) Don't count if bfqg complete all the requests.
>>
>> With the above changes, concurrent sync io can be supported if only
>> one group is activated.
>>
>> fio test script(startdelay is used to avoid queue merging):
>> [global]
>> filename=/dev/sda
>> allow_mounted_write=0
>> ioengine=psync
>> direct=1
>> ioscheduler=bfq
>> offset_increment=10g
>> group_reporting
>> rw=randwrite
>> bs=4k
>>
>> [test1]
>> numjobs=1
>>
>> [test2]
>> startdelay=1
>> numjobs=1
>>
>> [test3]
>> startdelay=2
>> numjobs=1
>>
>> [test4]
>> startdelay=3
>> numjobs=1
>>
>> [test5]
>> startdelay=4
>> numjobs=1
>>
>> [test6]
>> startdelay=5
>> numjobs=1
>>
>> [test7]
>> startdelay=6
>> numjobs=1
>>
>> [test8]
>> startdelay=7
>> numjobs=1
>>
>> test result:
>> running fio on root cgroup
>> v5.18:	   112 Mib/s
>> v5.18-patched: 112 Mib/s
>>
>> running fio on non-root cgroup
>> v5.18:	   51.2 Mib/s
>> v5.18-patched: 112 Mib/s
>>
>> Note that I also test null_blk with "irqmode=2
>> completion_nsec=100000000(100ms) hw_queue_depth=1", and tests show
>> that service guarantees are still preserved.
>>
> 
> Your patches seem ok to me now (thanks for you contribution and, above all, for your patience). I have only a high-level concern: what do you mean when you say that service guarantees are still preserved? What test did you run exactly? This point is very important to me. I'd like to see some convincing test with differentiated weights. In case you don't have other tools for executing such tests quickly, you may want to use the bandwidth-latency test in my simple S benchmark suite (for which I'm willing to help).

I'm runnnig some tests manually, just issuing same io to two cgroups,
and changing weights manually, specifically (1:10, 2:8, ..., 5:5),
then observe bandwidth from two cgroups.

Of course I'm glad to try your benchmark suite.

Thanks,
Kuai
> 
> Thanks,
> Paolo
> 
>> Previous versions:
>> RFC: https://lore.kernel.org/all/20211127101132.486806-1-yukuai3@huawei.com/
>> v1: https://lore.kernel.org/all/20220305091205.4188398-1-yukuai3@huawei.com/
>> v2: https://lore.kernel.org/all/20220416093753.3054696-1-yukuai3@huawei.com/
>> v3: https://lore.kernel.org/all/20220427124722.48465-1-yukuai3@huawei.com/
>> v4: https://lore.kernel.org/all/20220428111907.3635820-1-yukuai3@huawei.com/
>> v5: https://lore.kernel.org/all/20220428120837.3737765-1-yukuai3@huawei.com/
>> v6: https://lore.kernel.org/all/20220523131818.2798712-1-yukuai3@huawei.com/
>> v7: https://lore.kernel.org/all/20220528095020.186970-1-yukuai3@huawei.com/
>>
>>
>> Yu Kuai (6):
>>   block, bfq: support to track if bfqq has pending requests
>>   block, bfq: record how many queues have pending requests
>>   block, bfq: refactor the counting of 'num_groups_with_pending_reqs'
>>   block, bfq: do not idle if only one group is activated
>>   block, bfq: cleanup bfq_weights_tree add/remove apis
>>   block, bfq: cleanup __bfq_weights_tree_remove()
>>
>> block/bfq-cgroup.c  | 10 +++++++
>> block/bfq-iosched.c | 71 +++++++--------------------------------------
>> block/bfq-iosched.h | 30 +++++++++----------
>> block/bfq-wf2q.c    | 69 ++++++++++++++++++++++++++-----------------
>> 4 files changed, 76 insertions(+), 104 deletions(-)
>>
>> -- 
>> 2.31.1
>>
> 
> .
> 

Powered by blists - more mailing lists