Re: [dm-devel] [RFC] block: Change the granularity of io ticks from ms to ns

From: Jens Axboe <axboe@kernel.dk>
To: Yu Kuai <yukuai1@huaweicloud.com>, Ming Lei <ming.lei@redhat.com>
Cc: nvdimm@lists.linux.dev, linux-kernel@vger.kernel.org,
	song@kernel.org, dm-devel@redhat.com, ira.weiny@intel.com,
	agk@redhat.com, drbd-dev@lists.linbit.com, dave.jiang@intel.com,
	christoph.boehmwalder@linbit.com, vishal.l.verma@intel.com,
	konrad.wilk@oracle.com, "yukuai \(C\)" <yukuai3@huawei.com>,
	kent.overstreet@gmail.com, ngupta@vflare.org, kch@nvidia.com,
	senozhatsky@chromium.org,
	Gulam Mohamed <gulam.mohamed@oracle.com>,
	snitzer@kernel.org, colyli@suse.de, linux-block@vger.kernel.org,
	linux-bcache@vger.kernel.org, dan.j.williams@intel.com,
	linux-raid@vger.kernel.org, martin.petersen@oracle.com,
	philipp.reisner@linbit.com, junxiao.bi@oracle.com,
	minchan@kernel.org, lars.ellenberg@linbit.com
Subject: Re: [dm-devel] [RFC] block: Change the granularity of io ticks from ms to ns
Date: Wed, 7 Dec 2022 10:22:09 -0700	[thread overview]
Message-ID: <b8deb6fa-8a09-c1af-278f-24e66afe367d@kernel.dk> (raw)
In-Reply-To: <aadfc6d2-ad04-279c-a1d6-7f634d0b2c99@huaweicloud.com>

On 12/7/22 6:09 AM, Yu Kuai wrote:
> Hi,
> 
> 在 2022/12/07 11:15, Ming Lei 写道:
>> On Wed, Dec 07, 2022 at 10:19:08AM +0800, Yu Kuai wrote:
>>> Hi,
>>>
>>> 在 2022/12/07 2:15, Gulam Mohamed 写道:
>>>> Use ktime to change the granularity of IO accounting in block layer from
>>>> milli-seconds to nano-seconds to get the proper latency values for the
>>>> devices whose latency is in micro-seconds. After changing the granularity
>>>> to nano-seconds the iostat command, which was showing incorrect values for
>>>> %util, is now showing correct values.
>>>
>>> This patch didn't correct the counting of io_ticks, just make the
>>> error accounting from jiffies(ms) to ns. The problem that util can be
>>> smaller or larger still exist.
>>
>> Agree.
>>
>>>
>>> However, I think this change make sense consider that error margin is
>>> much smaller, and performance overhead should be minimum.
>>>
>>> Hi, Ming, how do you think?
>>
>> I remembered that ktime_get() has non-negligible overhead, is there any
>> test data(iops/cpu utilization) when running fio or t/io_uring on
>> null_blk with this patch?
> 
> Yes, testing with null_blk is necessary, we don't want any performance
> regression.

null_blk is fine as a substitute, but I'd much rather run this on my
test bench with actual IO and devices.

> BTW, I thought it's fine because it's already used for tracking io
> latency.

Reading a nsec timestamp is a LOT more expensive than reading jiffies,
which is essentially free. If you look at the amount of work that's
gone into minimizing ktime_get() for the fast path in the IO stack,
then that's a testament to that.

So that's a very bad assumption, and definitely wrong.

-- 
Jens Axboe

--
dm-devel mailing list
dm-devel@redhat.com
https://listman.redhat.com/mailman/listinfo/dm-devel