Re: [PATCH] xfs: account extra freespace btree splits for multiple allocations

From: Gao Xiang <hsiangkao@linux.alibaba.com>
To: Oliver Sang <oliver.sang@intel.com>,
	oe-lkp@lists.linux.dev, lkp@intel.com,
	Zirong Lang <zlang@redhat.com>,
	linux-xfs@vger.kernel.org, ying.huang@intel.com,
	feng.tang@intel.com, zhengjun.xing@linux.intel.com,
	fengwei.yin@intel.com, "Darrick J. Wong" <djwong@kernel.org>,
	Dave Chinner <dchinner@redhat.com>,
	Brian Foster <bfoster@redhat.com>,
	LKML <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH] xfs: account extra freespace btree splits for multiple allocations
Date: Wed, 23 Nov 2022 08:38:33 +0800	[thread overview]
Message-ID: <Y31riSRTC6P11bZt@B-P7TQMD6M-0146.local> (raw)
In-Reply-To: <Y31qPk6+h80cFGXn@B-P7TQMD6M-0146.local>

Sorry I sent out a staging reply, please ignore this.

On Wed, Nov 23, 2022 at 08:33:02AM +0800, Gao Xiang wrote:
> On Tue, Nov 22, 2022 at 06:03:03PM +0800, Oliver Sang wrote:
> > hi Gao Xiang,
> > 
> > On Tue, Nov 22, 2022 at 09:33:38AM +0800, Gao Xiang wrote:
> > > On Tue, Nov 22, 2022 at 09:09:34AM +0800, kernel test robot wrote:
> > > > 
> > > > please be noted we noticed Gao Xiang and Dave Chinner have already had lots of
> > > > discussion around this patch, which seems there is maybe new version later.
> > > > we just sent out this report FYI the possible performance impact of this patch.
> > > > 
> > > > 
> > > > Greeting,
> > > > 
> > > > FYI, we noticed a -15.1% regression of fxmark.ssd_xfs_MWCM_72_directio.works/sec due to commit:
> > > 
> > > Thanks for your report!
> > > 
> > > At a glance, I have no idea why this commit can have performance
> > > impacts.  Is the result stable?
> > 
> > in our tests, the result is quite stable.
> >      45589           -15.1%      38687 ±  2%  fxmark.ssd_xfs_MWCM_72_directio.works/sec
> > 
> > and detail data is as below:
> > for this commit:
> >   "fxmark.ssd_xfs_MWCM_72_directio.works/sec": [
> >     39192.224368,
> >     39665.690567,
> >     38980.680601,
> >     37298.99538,
> >     37483.256377,
> >     39504.606569
> >   ],
> > 
> > for parent:
> >   "fxmark.ssd_xfs_MWCM_72_directio.works/sec": [
> >     45381.458009,
> >     45314.376204,
> >     45724.688965,
> >     45751.955937,
> >     45614.323267,
> >     45747.216475
> >   ],
> > 
> > 
> > if you still have concern, we could rerun tests. Thanks!
> 
> According to the report, I can see:
> 	67262           -29.6%      47384 ±  7%  fxmark.ssd_xfs_MWCM_18_directio.works/sec
> 	52786           -15.6%      44567 ±  4%  fxmark.ssd_xfs_MWCM_1_directio.works/sec
> 	63189           -23.3%      48486 ±  5%  fxmark.ssd_xfs_MWCM_2_directio.works/sec

I meant allocation strategy change may cause this, but I didn't think it
could cause such impact.  I will reconfirm on my own side as well.

Thanks,
Gao Xiang