From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail-pb0-x236.google.com (mail-pb0-x236.google.com [IPv6:2607:f8b0:400e:c01::236]) by dpdk.org (Postfix) with ESMTP id EDBFD30E for ; Tue, 28 May 2013 06:00:19 +0200 (CEST) Received: by mail-pb0-f54.google.com with SMTP id ro12so7340753pbb.41 for ; Mon, 27 May 2013 21:00:25 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20120113; h=mime-version:in-reply-to:references:date:message-id:subject:from:to :cc:content-type; bh=UCwugcjD5ZvrJgFo+27+JP3GAMul86FxYrkP/7SxzvQ=; b=PRJdNT2z+o7fnFdy5phC5Lg5kDaFTlvo6NBcyVwWxCMU1voS8puTow36ADPRUkStKy mw2FdCA1Z6rgKLgQVMlrrig7PEJCmLYocMtV/rHTxJt2p/xfQPZ5sjRzBaaf8UBg0Pqt vJmF8xEMyQNg/fR5MdMAMgXgiXneD6sXyjhzkGwyjFyQW70IW8z02qxDq9aKBkxKucud 9280Gd9xJwFta4bP19OHbd6fpB+Z989GYOlhIofvvSWOlO9rj8OijW7kIDWsqZxFsA39 3Bp9pV8ni0rc3P7/ic+J/T8hCixx2xg3LYEZq7LA4vJqLuXw/OMmNZ2NwXyoqn6j/ti2 5XVQ== MIME-Version: 1.0 X-Received: by 10.68.224.104 with SMTP id rb8mr31872299pbc.206.1369713625402; Mon, 27 May 2013 21:00:25 -0700 (PDT) Received: by 10.66.155.170 with HTTP; Mon, 27 May 2013 21:00:25 -0700 (PDT) In-Reply-To: <20130527202946.75184090@nehalam.linuxnetplumber.net> References: <20130528113005.5E6D.C42C3789@sakura.ad.jp> <51A4214B.8040703@6wind.com> <20130527202946.75184090@nehalam.linuxnetplumber.net> Date: Tue, 28 May 2013 13:00:25 +0900 Message-ID: From: Shinae Woo To: Stephen Hemminger Content-Type: multipart/alternative; boundary=047d7b1602a116123104ddbf4e20 Cc: dev@dpdk.org Subject: Re: [dpdk-dev] Performances are not scale with multiple ports X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: patches and discussions about DPDK List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Tue, 28 May 2013 04:00:20 -0000 --047d7b1602a116123104ddbf4e20 Content-Type: text/plain; charset=ISO-8859-1 Actually, I made my own packet reception application using DPDK to utilize multiple RSS which map each core (in my case, 12 queues for 12 cores). But it shows that multiple RSS shows, the reception performance is rather decreasing as below. # queue 1 2 3 4 5 6 Bandwidth (Gbps) 8.95 8.90 8.89 8.57 7.95 7.62 #queue 7 8 9 10 11 12 Bandwidth (Gbps) 7.18 7.13 7.12 7.04 7.58 6.80 I will try with testpmd to load balance again. I use the custom packet generator module over Packet I/O, but I'm looking forward to the pktgen for DPDK of Stephen's, too. Thanks, Shinae On Tue, May 28, 2013 at 12:29 PM, Stephen Hemminger < stephen@networkplumber.org> wrote: > On Mon, 27 May 2013 20:15:23 -0700 > Emre Eraltan wrote: > > > Hello Shinae, > > > > Did you try to use the testpmd tool with multiple queues per port? It > > gives you more flexibility compared to l2fwd app. > > > > You need to trigger the RSS feature of the NIC by sending different > > streams (just by changing the destination port for instance or any > > information in the 5-tuple). This will load balance your packets among > > several cores so that you can probe multiple queues with different > > cores. Otherwise, you will use only one core (or thread if HT is > > enabled) per port for the RX side. > > > > Best Regards, > > Emre > > > > I wrote a partial baked version of pktgen for DPDK. Let me see if I can > get ok to release and clean it up. > > Windriver has one as well, and they use it in their demo clases. > Not sure why it never got included. > > --047d7b1602a116123104ddbf4e20 Content-Type: text/html; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable

Actually, I made my own packet receptio= n application using DPDK
to utilize multiple RSS which map each core (i= n my case, 12 queues for 12 cores).

But it shows that multiple= RSS shows, the reception performance is rather decreasing as below.

<= tbody>
# queue
 1 2 3 4 5 6
Bandwidth (Gbps)=A0 8.95 8.90 8.89 8.57 7.95 7.62
#= queue 7 8 9 10 11 12
Bandwidth (Gbps)=A0 7.18 7.13 7.12 7.04 7.58 6.80

I will try wi= th testpmd to load balance again.

I use the cu= stom packet generator module over Packet I/O,
but I'm loo= king forward to the pktgen for DPDK of Stephen's, too.

Thanks,
Shinae

<= div class=3D"gmail_extra">

On Tue, May 28= , 2013 at 12:29 PM, Stephen Hemminger <stephen@networkplumber.org= > wrote:
On Mon, 27 May 2013 20:15:= 23 -0700
Emre Eraltan <emre.eraltan@6wi= nd.com> wrote:

> Hello Shinae,
>
> Did you try to use the testpmd tool with multiple queues per port? It<= br> > gives you more flexibility compared to l2fwd app.
>
> You need to trigger the RSS feature of the NIC by sending different > streams (just by changing the destination port for instance or any
> information in the 5-tuple). This will load balance your packets among=
> several cores so that you can probe multiple queues with different
> cores. Otherwise, you will use only one core (or thread if HT is
> enabled) per port for the RX side.
>
> Best Regards,
> Emre
>

I wrote a partial baked version of pktgen for DPDK. Let me see if I c= an
get ok to release and clean it up.

Windriver has one as well, and they use it in their demo clases.
Not sure why it never got included.


--047d7b1602a116123104ddbf4e20--