Received-SPF: pass (google.com: domain of linux-kernel+bounces-204842-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45d1:ec00::1 as permitted sender) client-ip=2604:1380:45d1:ec00::1;
Precedence: bulk
MIME-Version: 1.0
References: <20240508202111.768b7a4d@yea> <20240515224524.1c8befbe@yea>
 <CAOUHufZ-9NmzOKjLedvZFp0=N0LvRZn77qC6k1WXK+NHtKr=0w@mail.gmail.com>
 <CAOUHufZ36rQc8AfLtRv2QrEareysdvbprAEO5XkcG-FeDOxFLA@mail.gmail.com>
 <20240602200332.3e531ff1@yea> <20240604001304.5420284f@yea>
 <CAJD7tkbCRLdy0vD2Pd17fNrxHgkzW1VucN4qMkohLFLBLaaeCQ@mail.gmail.com>
 <20240604134458.3ae4396a@yea> <CAJD7tkYjJJGthQ_8NukGw6Q9EYbLA=8sAH_7=B90KXEL6HWdSw@mail.gmail.com>
 <CAOUHufa0Fpj6SjNgB-z0n5Jg63q1ewkbOAU65forpDwQVs45qg@mail.gmail.com>
 <CAJD7tkb=5GJ9SNUwDsu1Zy3Tus4rjsNo60Hg9N7=gGth409Diw@mail.gmail.com>
 <CAOUHufb6zXr14Wm3T-4-OJh7iAq+vzDKwVYfHLhMMt96SpiZXg@mail.gmail.com>
 <CAJD7tkZ+QY55GTzW9A7ZCm=rxAEfrW76cWXf8o5nwiKSXp8z=w@mail.gmail.com>
 <20240604231019.18e2f373@yea> <CAJD7tkYq5u7B+0UH2XKpeWJnUxoO2kJ1_XZ2JOgYpyNEVR7u0g@mail.gmail.com>
 <20240606010431.2b33318c@yea> <CAJD7tkbhWYzx=6YmzAh0F+cK-_Bn8mPOH7gMbQS7YVXmaFSgFg@mail.gmail.com>
 <CAPpoddc2vOLdQJ7HwG7x+=oZsTz221+YJcNbUtKvPjA9AyeY2w@mail.gmail.com> <CAJD7tkb2b0+4_m0gb8DKSTtRwtC2GMa9NF5RuGKhXJARYHK0gw@mail.gmail.com>
In-Reply-To: <CAJD7tkb2b0+4_m0gb8DKSTtRwtC2GMa9NF5RuGKhXJARYHK0gw@mail.gmail.com>
From: Yu Zhao <yuzhao@google.com>
Date: Thu, 6 Jun 2024 11:55:00 -0600
Message-ID: <CAOUHufZ_dKpts4uW4Xg3jVYnX5Z5MN9U9icxhP0Nmz7QnqurBQ@mail.gmail.com>
Subject: Re: kswapd0: page allocation failure: order:0, mode:0x820(GFP_ATOMIC),
 nodemask=(null),cpuset=/,mems_allowed=0 (Kernel v6.5.9, 32bit ppc)
To: Yosry Ahmed <yosryahmed@google.com>, Takero Funaki <flintglass@gmail.com>
Cc: Erhard Furtner <erhard_f@mailbox.org>, linux-mm@kvack.org, linux-kernel@vger.kernel.org, 
	linuxppc-dev@lists.ozlabs.org, Johannes Weiner <hannes@cmpxchg.org>, 
	Nhat Pham <nphamcs@gmail.com>, Chengming Zhou <chengming.zhou@linux.dev>, 
	Sergey Senozhatsky <senozhatsky@chromium.org>, Minchan Kim <minchan@kernel.org>, 
	"Vlastimil Babka (SUSE)" <vbabka@kernel.org>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Thu, Jun 6, 2024 at 11:42=E2=80=AFAM Yosry Ahmed <yosryahmed@google.com>=
 wrote:
>
> On Thu, Jun 6, 2024 at 10:14=E2=80=AFAM Takero Funaki <flintglass@gmail.c=
om> wrote:
> >
> > 2024=E5=B9=B46=E6=9C=886=E6=97=A5(=E6=9C=A8) 8:42 Yosry Ahmed <yosryahm=
ed@google.com>:
> >
> > > I think there are multiple ways to go forward here:
> > > (a) Make the number of zpools a config option, leave the default as
> > > 32, but allow special use cases to set it to 1 or similar. This is
> > > probably not preferable because it is not clear to users how to set
> > > it, but the idea is that no one will have to set it except special us=
e
> > > cases such as Erhard's (who will want to set it to 1 in this case).
> > >
> > > (b) Make the number of zpools scale linearly with the number of CPUs.
> > > Maybe something like nr_cpus/4 or nr_cpus/8. The problem with this
> > > approach is that with a large number of CPUs, too many zpools will
> > > start having diminishing returns. Fragmentation will keep increasing,
> > > while the scalability/concurrency gains will diminish.
> > >
> > > (c) Make the number of zpools scale logarithmically with the number o=
f
> > > CPUs. Maybe something like 4log2(nr_cpus). This will keep the number
> > > of zpools from increasing too much and close to the status quo. The
> > > problem is that at a small number of CPUs (e.g. 2), 4log2(nr_cpus)
> > > will actually give a nr_zpools > nr_cpus. So we will need to come up
> > > with a more fancy magic equation (e.g. 4log2(nr_cpus/4)).
> > >
> >
> > I just posted a patch to limit the number of zpools, with some
> > theoretical background explained in the code comments. I believe that
> > 2 * CPU linearly is sufficient to reduce contention, but the scale can
> > be reduced further. All CPUs are trying to allocate/free zswap is
> > unlikely to happen.
> >  How many concurrent accesses were the original 32 zpools supposed to
> > handle? I think it was for 16 cpu or more. or nr_cpus/4 would be
> > enough?
>
> We use 32 zpools on machines with 100s of CPUs. Two zpools per CPU is
> an overkill imo.

Not to choose a camp; just a friendly note on why I strongly disagree
with the N zpools per CPU approach:
1. It is fundamentally flawed to assume the system is linear;
2. Nonlinear systems usually have diminishing returns.

For Google data centers, using nr_cpus as the scaling factor had long
passed the acceptable ROI threshold. Per-CPU data, especially when
compounded per memcg or even per process, is probably the number-one
overhead in terms of DRAM efficiency.


> I have further comments that I will leave on the patch, but I mainly
> think this should be driven by real data, not theoretical possibility
> of lock contention.