From: Paolo Abeni <pabeni@redhat.com>
To: Eric Dumazet <eric.dumazet@gmail.com>
Cc: David Miller <davem@davemloft.net>,
netdev <netdev@vger.kernel.org>,
Willem de Bruijn <willemb@google.com>,
Tom Herbert <tom@herbertland.com>
Subject: Re: [PATCH net-next] net: sock_rps_record_flow() is for connected sockets
Date: Thu, 08 Dec 2016 18:49:58 +0100 [thread overview]
Message-ID: <1481219398.6120.1.camel@redhat.com> (raw)
In-Reply-To: <1481120791.4930.4.camel@edumazet-glaptop3.roam.corp.google.com>
On Wed, 2016-12-07 at 06:26 -0800, Eric Dumazet wrote:
> On Wed, 2016-12-07 at 08:57 +0100, Paolo Abeni wrote:
> > On Tue, 2016-12-06 at 22:47 -0800, Eric Dumazet wrote:
> > > On Tue, 2016-12-06 at 19:32 -0800, Eric Dumazet wrote:
> > > > A follow up patch will provide a static_key (Jump Label) since most
> > > > hosts do not even use RFS.
> > >
> > > Speaking of static_key, it appears we now have GRO on UDP, and this
> > > consumes a considerable amount of cpu cycles.
> > >
> > > Turning off GRO allows me to get +20 % more packets on my single UDP
> > > socket. (1.2 Mpps instead of 1.0 Mpps)
> >
> > I see also an improvement for single flow tests disabling GRO, but on a
> > smaller scale (~5% if I recall correctly).
>
> Was it on a NUMA host ?
I'm using a single socket host, with 12 cores/24 threads and 16 RX
queues.
But my data is old. I'll re-run the test on top of current net-next.
Paolo
next prev parent reply other threads:[~2016-12-08 17:50 UTC|newest]
Thread overview: 43+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-12-05 2:43 [RFC] udp: some improvements on RX path Eric Dumazet
2016-12-05 13:22 ` Paolo Abeni
2016-12-05 14:28 ` Eric Dumazet
2016-12-05 15:37 ` Jesper Dangaard Brouer
2016-12-05 15:54 ` Eric Dumazet
2016-12-05 17:57 ` [PATCH] net/udp: do not touch skb->peeked unless really needed Eric Dumazet
2016-12-06 9:53 ` Paolo Abeni
2016-12-06 12:10 ` Paolo Abeni
2016-12-06 14:35 ` Eric Dumazet
2016-12-06 14:34 ` Eric Dumazet
2016-12-06 10:34 ` Paolo Abeni
2016-12-06 17:08 ` Paolo Abeni
2016-12-06 17:47 ` Eric Dumazet
2016-12-06 18:31 ` Paolo Abeni
2016-12-06 18:58 ` Eric Dumazet
2016-12-06 19:16 ` Paolo Abeni
2016-12-06 19:35 ` Eric Dumazet
2016-12-07 3:32 ` [PATCH net-next] net: sock_rps_record_flow() is for connected sockets Eric Dumazet
2016-12-07 6:47 ` Eric Dumazet
2016-12-07 7:57 ` Paolo Abeni
2016-12-07 14:26 ` Eric Dumazet
2016-12-08 17:49 ` Paolo Abeni [this message]
2016-12-07 14:29 ` Eric Dumazet
2016-12-07 15:59 ` Eric Dumazet
2016-12-08 18:50 ` Paolo Abeni
2016-12-08 19:32 ` Eric Dumazet
2016-12-08 19:20 ` Edward Cree
2016-12-08 17:49 ` Tom Herbert
2016-12-08 18:02 ` Eric Dumazet
2016-12-08 19:15 ` Tom Herbert
2016-12-08 20:05 ` Hannes Frederic Sowa
2016-12-08 20:30 ` Tom Herbert
2016-12-08 20:44 ` Tom Herbert
2016-12-08 18:07 ` Eric Dumazet
2016-12-07 7:59 ` Paolo Abeni
2016-12-07 13:58 ` Eric Dumazet
2016-12-07 15:47 ` David Miller
2016-12-07 17:09 ` [PATCH] net/udp: do not touch skb->peeked unless really needed David Laight
2016-12-07 17:32 ` Eric Dumazet
2016-12-07 17:37 ` Hannes Frederic Sowa
2016-12-07 17:52 ` Eric Dumazet
2016-12-07 17:55 ` Eric Dumazet
2016-12-06 15:42 ` David Miller
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 \
--in-reply-to=1481219398.6120.1.camel@redhat.com \
--to=pabeni@redhat.com \
--cc=davem@davemloft.net \
--cc=eric.dumazet@gmail.com \
--cc=netdev@vger.kernel.org \
--cc=tom@herbertland.com \
--cc=willemb@google.com \
/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;
as well as URLs for NNTP newsgroup(s).