From: Pavel Begunkov <[email protected]>
To: David Ahern <[email protected]>,
[email protected], [email protected],
[email protected]
Cc: "David S . Miller" <[email protected]>,
Jakub Kicinski <[email protected]>,
Jonathan Lemon <[email protected]>,
Willem de Bruijn <[email protected]>,
Jens Axboe <[email protected]>,
[email protected]
Subject: Re: [PATCH net-next v4 00/27] io_uring zerocopy send
Date: Fri, 8 Jul 2022 15:26:13 +0100 [thread overview]
Message-ID: <[email protected]> (raw)
In-Reply-To: <[email protected]>
On 7/8/22 05:10, David Ahern wrote:
> On 7/7/22 5:49 AM, Pavel Begunkov wrote:
>> NOTE: Not be picked directly. After getting necessary acks, I'll be working
>> out merging with Jakub and Jens.
>>
>> The patchset implements io_uring zerocopy send. It works with both registered
>> and normal buffers, mixing is allowed but not recommended. Apart from usual
>> request completions, just as with MSG_ZEROCOPY, io_uring separately notifies
>> the userspace when buffers are freed and can be reused (see API design below),
>> which is delivered into io_uring's Completion Queue. Those "buffer-free"
>> notifications are not necessarily per request, but the userspace has control
>> over it and should explicitly attaching a number of requests to a single
>> notification. The series also adds some internal optimisations when used with
>> registered buffers like removing page referencing.
>>
>> From the kernel networking perspective there are two main changes. The first
>> one is passing ubuf_info into the network layer from io_uring (inside of an
>> in kernel struct msghdr). This allows extra optimisations, e.g. ubuf_info
>> caching on the io_uring side, but also helps to avoid cross-referencing
>> and synchronisation problems. The second part is an optional optimisation
>> removing page referencing for requests with registered buffers.
>>
>> Benchmarking with an optimised version of the selftest (see [1]), which sends
>> a bunch of requests, waits for completions and repeats. "+ flush" column posts
>> one additional "buffer-free" notification per request, and just "zc" doesn't
>> post buffer notifications at all.
>>
>> NIC (requests / second):
>> IO size | non-zc | zc | zc + flush
>> 4000 | 495134 | 606420 (+22%) | 558971 (+12%)
>> 1500 | 551808 | 577116 (+4.5%) | 565803 (+2.5%)
>> 1000 | 584677 | 592088 (+1.2%) | 560885 (-4%)
>> 600 | 596292 | 598550 (+0.4%) | 555366 (-6.7%)
>>
>> dummy (requests / second):
>> IO size | non-zc | zc | zc + flush
>> 8000 | 1299916 | 2396600 (+84%) | 2224219 (+71%)
>> 4000 | 1869230 | 2344146 (+25%) | 2170069 (+16%)
>> 1200 | 2071617 | 2361960 (+14%) | 2203052 (+6%)
>> 600 | 2106794 | 2381527 (+13%) | 2195295 (+4%)
>>
>> Previously it also brought a massive performance speedup compared to the
>> msg_zerocopy tool (see [3]), which is probably not super interesting.
>>
>
> can you add a comment that the above results are for UDP.
Oh, right, forgot to add it
> You dropped comments about TCP testing; any progress there? If not, can
> you relay any issues you are hitting?
Not really a problem, but for me it's bottle necked at NIC bandwidth
(~3GB/s) for both zc and non-zc and doesn't even nearly saturate a CPU.
Was actually benchmarked by my colleague quite a while ago, but can't
find numbers. Probably need to at least add localhost numbers or grab
a better server.
--
Pavel Begunkov
next prev parent reply other threads:[~2022-07-08 14:28 UTC|newest]
Thread overview: 45+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-07-07 11:49 [PATCH net-next v4 00/27] io_uring zerocopy send Pavel Begunkov
2022-07-07 11:49 ` [PATCH net-next v4 01/27] ipv4: avoid partial copy for zc Pavel Begunkov
2022-07-07 11:49 ` [PATCH net-next v4 02/27] ipv6: " Pavel Begunkov
2022-07-07 11:49 ` [PATCH net-next v4 03/27] skbuff: don't mix ubuf_info from different sources Pavel Begunkov
2022-07-07 11:49 ` [PATCH net-next v4 04/27] skbuff: add SKBFL_DONT_ORPHAN flag Pavel Begunkov
2022-07-07 11:49 ` [PATCH net-next v4 05/27] skbuff: carry external ubuf_info in msghdr Pavel Begunkov
2022-07-07 11:49 ` [PATCH net-next v4 06/27] net: Allow custom iter handler " Pavel Begunkov
2022-07-11 12:20 ` Pavel Begunkov
2022-07-07 11:49 ` [PATCH net-next v4 07/27] net: introduce managed frags infrastructure Pavel Begunkov
2022-07-07 11:49 ` [PATCH net-next v4 08/27] net: introduce __skb_fill_page_desc_noacc Pavel Begunkov
2022-07-07 11:49 ` [PATCH net-next v4 09/27] ipv4/udp: support externally provided ubufs Pavel Begunkov
2022-07-07 11:49 ` [PATCH net-next v4 10/27] ipv6/udp: " Pavel Begunkov
2022-07-07 11:49 ` [PATCH net-next v4 11/27] tcp: " Pavel Begunkov
2022-07-08 4:06 ` David Ahern
2022-07-08 14:03 ` Pavel Begunkov
2022-07-13 23:38 ` David Ahern
2022-07-07 11:49 ` [PATCH net-next v4 12/27] io_uring: initialise msghdr::msg_ubuf Pavel Begunkov
2022-07-07 11:49 ` [PATCH net-next v4 13/27] io_uring: export io_put_task() Pavel Begunkov
2022-07-07 11:49 ` [PATCH net-next v4 14/27] io_uring: add zc notification infrastructure Pavel Begunkov
2022-07-07 11:49 ` [PATCH net-next v4 15/27] io_uring: cache struct io_notif Pavel Begunkov
2022-07-07 11:49 ` [PATCH net-next v4 16/27] io_uring: complete notifiers in tw Pavel Begunkov
2022-07-07 11:49 ` [PATCH net-next v4 17/27] io_uring: add rsrc referencing for notifiers Pavel Begunkov
2022-07-07 11:49 ` [PATCH net-next v4 18/27] io_uring: add notification slot registration Pavel Begunkov
2022-07-07 11:49 ` [PATCH net-next v4 19/27] io_uring: wire send zc request type Pavel Begunkov
2022-07-07 11:49 ` [PATCH net-next v4 20/27] io_uring: account locked pages for non-fixed zc Pavel Begunkov
2022-07-07 11:49 ` [PATCH net-next v4 21/27] io_uring: allow to pass addr into sendzc Pavel Begunkov
2022-07-07 11:49 ` [PATCH net-next v4 22/27] io_uring: sendzc with fixed buffers Pavel Begunkov
2022-07-07 11:49 ` [PATCH net-next v4 23/27] io_uring: flush notifiers after sendzc Pavel Begunkov
2022-07-07 11:49 ` [PATCH net-next v4 24/27] io_uring: rename IORING_OP_FILES_UPDATE Pavel Begunkov
2022-07-07 11:49 ` [PATCH net-next v4 25/27] io_uring: add zc notification flush requests Pavel Begunkov
2022-07-07 11:49 ` [PATCH net-next v4 26/27] io_uring: enable managed frags with register buffers Pavel Begunkov
2022-07-07 11:49 ` [PATCH net-next v4 27/27] selftests/io_uring: test zerocopy send Pavel Begunkov
2022-07-08 4:10 ` [PATCH net-next v4 00/27] io_uring " David Ahern
2022-07-08 14:26 ` Pavel Begunkov [this message]
2022-07-11 12:56 ` Pavel Begunkov
2022-07-13 23:45 ` David Ahern
2022-07-14 18:55 ` Pavel Begunkov
2022-07-18 2:19 ` David Ahern
2022-07-20 13:32 ` Pavel Begunkov
2022-07-24 18:28 ` David Ahern
2022-07-27 10:51 ` Pavel Begunkov
2022-07-29 22:30 ` David Ahern
2022-09-26 20:08 ` Pavel Begunkov
2022-09-28 19:31 ` David Ahern
2022-09-28 20:11 ` Pavel Begunkov
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
[email protected] \
[email protected] \
[email protected] \
[email protected] \
[email protected] \
[email protected] \
[email protected] \
[email protected] \
[email protected] \
[email protected] \
[email protected] \
[email protected] \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox