Feedback-ID: i97614980:Fastmail
MIME-Version: 1.0
Date: Sat, 26 Jul 2025 03:16:24 +0200
From: "Pierre Barre" <pierre@barre.sh>
To: "Jeff Ross" <jross@openvistas.net>, pgsql-general@lists.postgresql.org
Message-Id: <f764be56-93ff-4062-a081-311b9b6cae43@app.fastmail.com>
In-Reply-To: <77eb549f-ef2d-46c1-932d-c54247e1400a@app.fastmail.com>
References: <a9fe5ddb-9685-4139-bc1f-88161a7a4da3@app.fastmail.com>
 <CAG1bHGOzCNtDeW0W8gRO7mpW=t7BqWh-iz4kX5VRCPgt_6Tr6Q@mail.gmail.com>
 <8188513c-e089-4273-b2be-16dd0a5a0a80@app.fastmail.com>
 <aIKNLuEis1DQEOGv@ubby>
 <5c512367-0f67-4bcc-9897-1acf9c0f8bd3@app.fastmail.com>
 <60027457-1b85-4a69-a67e-ee87f7cabd61@openvistas.net>
 <b62bd887-bbb8-4ade-8947-9d2608618b02@app.fastmail.com>
 <77eb549f-ef2d-46c1-932d-c54247e1400a@app.fastmail.com>
Subject: Re: PostgreSQL on S3-backed Block Storage with Near-Local Performance
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: quoted-printable
Archived-At: <https://www.postgresql.org/message-id/f764be56-93ff-4062-a081-311b9b6cae43%40app.fastmail.com>
Precedence: bulk

I built postgres (same version, 16.9) but --with-block-size=3D32 (I'd re=
ally love if this would be a initdb time flag!) and did some more testin=
g:

synchronous_commit =3D off

postgres@zerofs:~$ pgbench -vvv -c 100 -j 40 -t 10000 bench
pgbench (16.9 (Ubuntu 16.10-1))
starting vacuum...end.
starting vacuum pgbench_accounts...end.
transaction type: <builtin: TPC-B (sort of)>
scaling factor: 50
query mode: simple
number of clients: 100
number of threads: 40
maximum number of tries: 1
number of transactions per client: 10000
number of transactions actually processed: 1000000/1000000
number of failed transactions: 0 (0.000%)
latency average =3D 5.727 ms
initial connection time =3D 59.223 ms
tps =3D 17460.128835 (without initial connection time)

synchronous_commit =3D on=20

postgres@zerofs:/root$ pgbench -vvv -c 100 -j 40 -t 1000 bench
pgbench (16.9 (Ubuntu 16.10-1))
starting vacuum...end.
starting vacuum pgbench_accounts...end.
transaction type: <builtin: TPC-B (sort of)>
scaling factor: 50
query mode: simple
number of clients: 100
number of threads: 40
maximum number of tries: 1
number of transactions per client: 1000
number of transactions actually processed: 100000/100000
number of failed transactions: 0 (0.000%)
latency average =3D 301.800 ms
initial connection time =3D 62.237 ms
tps =3D 331.345391 (without initial connection time)

=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D

Then, using the same setup (same server, same postgres build), I create =
a ZeroFS NBD device with ext4 on top

/dev/nbd0 on /mnt_9p type ext4 (rw,relatime,stripe=3D32)

synchronous_commit =3D off

postgres@zerofs:/mnt_9p$ pgbench -vvv -c 100 -j 40 -t 10000 bench
pgbench (16.9 (Ubuntu 16.10-1))
starting vacuum...end.
starting vacuum pgbench_accounts...end.
transaction type: <builtin: TPC-B (sort of)>
scaling factor: 50
query mode: simple
number of clients: 100
number of threads: 40
maximum number of tries: 1
number of transactions per client: 10000
number of transactions actually processed: 1000000/1000000
number of failed transactions: 0 (0.000%)
latency average =3D 3.615 ms
initial connection time =3D 45.653 ms
tps =3D 27665.373366 (without initial connection time)

synchronous_commit =3D on

postgres@zerofs:/root$ pgbench -vvv -c 100 -j 40 -t 1000 bench
pgbench (16.9 (Ubuntu 16.10-1))
starting vacuum...end.
starting vacuum pgbench_accounts...end.
transaction type: <builtin: TPC-B (sort of)>
scaling factor: 50
query mode: simple
number of clients: 100
number of threads: 40
maximum number of tries: 1
number of transactions per client: 1000
number of transactions actually processed: 100000/100000
number of failed transactions: 0 (0.000%)
latency average =3D 337.762 ms
initial connection time =3D 43.969 ms
tps =3D 296.066616 (without initial connection time)

Best,
Pierre


On Fri, Jul 25, 2025, at 11:25, Pierre Barre wrote:
> Hi,
>
> I went ahead and did that test.
>
> Here is the postgresql config I used for reference (note the wal=20
> options (recycle, init_zero) as well as full_page_writes =3D off, beca=
use=20
> ZeroFS cannot have torn writes by design).
>
> https://gist.github.com/Barre/8d68f0d00446389998a31f4e60f3276d
>
> Test was running on Azure with Standard D16ads v5 (16 vcpus, 64 GiB me=
mory)
>
> This time, I didn't run ZFS with L2ARC, I just mounted ZeroFS with 9p.
>
> synchronous_commit =3D off=20
>
> postgres@zerofs:~$ pgbench -vvv -c 100 -j 40 -t 1000 bench
> pgbench (16.9 (Ubuntu 16.9-0ubuntu0.24.04.1))
> starting vacuum...end.
> starting vacuum pgbench_accounts...end.
> transaction type: <builtin: TPC-B (sort of)>
> scaling factor: 50
> query mode: simple
> number of clients: 100
> number of threads: 40
> maximum number of tries: 1
> number of transactions per client: 1000
> number of transactions actually processed: 100000/100000
> number of failed transactions: 0 (0.000%)
> latency average =3D 6.239 ms
> initial connection time =3D 68.922 ms
> tps =3D 16026.940646 (without initial connection time)
>
>
> synchronous_commit =3D on
>
> postgres@zerofs:~$ pgbench -vvv -c 50 -j 15 -t 1000 bench
> pgbench (16.9 (Ubuntu 16.9-0ubuntu0.24.04.1))
> starting vacuum...end.
> starting vacuum pgbench_accounts...end.
> transaction type: <builtin: TPC-B (sort of)>
> scaling factor: 50
> query mode: simple
> number of clients: 50
> number of threads: 15
> maximum number of tries: 1
> number of transactions per client: 1000
> number of transactions actually processed: 50000/50000
> number of failed transactions: 0 (0.000%)
> latency average =3D 197.723 ms
> initial connection time =3D 46.089 ms
> tps =3D 252.878721 (without initial connection time)
>
>
> Not great barebones with with synchronous_commit, but still usable!
>
> Best,
> Pierre
>
> On Fri, Jul 25, 2025, at 00:44, Pierre Barre wrote:
>>> This then begs the obvious question of how fast is this with=20
>>> synchronous_commit =3D on?
>>
>> Probably not awful, especially with commit_delay.
>>
>> I'll try that and report back.
>>
>> Best,
>> Pierre
>>
>> On Fri, Jul 25, 2025, at 00:03, Jeff Ross wrote:
>>> On 7/24/25 13:50, Pierre Barre wrote:
>>>
>>>> It=E2=80=99s not =E2=80=9Csafe=E2=80=9D or =E2=80=9Cunsafe=E2=80=9D=
, there=E2=80=99s mountains of valid workloads which don=E2=80=99t requi=
re synchronous_commit. Synchronous_commit don=E2=80=99t make your system=
 automatically safe either, and if that=E2=80=99s a requirement, there=E2=
=80=99s many workarounds, as you suggested, it certainly doesn=E2=80=99t=
 make the setup useless.
>>>>
>>>> Best,
>>>> Pierre
>>>>
>>>> On Thu, Jul 24, 2025, at 21:44, Nico Williams wrote:
>>>>> On Fri, Jul 18, 2025 at 12:57:39PM +0200, Pierre Barre wrote:
>>>>>> - Postgres configured accordingly memory-wise as well as with
>>>>>>    synchronous_commit =3D off, wal_init_zero =3D off and wal_recy=
cle =3D off.
>>>>> Bingo.  That's why it's fast (synchronous_commit =3D off).  It's a=
lso why
>>>>> it's not safe _unless_ you have a local, fast, persistent ZIL devi=
ce
>>>>> (which I assume you don't).
>>>>>
>>>>> Nico
>>>>> --
>>> This then begs the obvious question of how fast is this with=20
>>> synchronous_commit =3D on?