All of lore.kernel.org
 help / color / mirror / Atom feed
From: "Ernesto Ruffini" <eruffini@outsys.org>
To: "'Ayshathul Thuhara'" <ayshathul.thuhara@iwave-global.com>
Cc: "'dev'" <dev@dpdk.org>, <yasu1976@gmail.com>, <users@dpdk.org>
Subject: RE: Guidance Required to Improve Throughput on ConnectX-7 (1x400G) with DPDK v23.07
Date: Fri, 13 Feb 2026 11:09:11 +0100	[thread overview]
Message-ID: <03ad01dc9cd0$cd2cfb00$6786f100$@outsys.org> (raw)
In-Reply-To: <PN0P287MB2531BFD1D63007182EFA3E5CC561A@PN0P287MB2531.INDP287.PROD.OUTLOOK.COM>

[-- Attachment #1: Type: text/plain, Size: 4133 bytes --]

Hi Ayshathul Thuhara,

l3fwd parameters in the doc are important: they say how many cores/queues to
use.

Try playing with them, adding testpmd specific parameters.

 

sudo ./build/app/dpdk-testpmd -a 01:00.0 -l
0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16 -- -i --rxq=16 --txq=16
--nb-cores=16

 

Regards,

               Ernesto

 

 

From: Ayshathul Thuhara <ayshathul.thuhara@iwave-global.com> 
Sent: Friday, February 13, 2026 10:04
To: Ernesto Ruffini <eruffini@outsys.org>
Cc: 'dev' <dev@dpdk.org>; yasu1976@gmail.com; users@dpdk.org
Subject: Re: Guidance Required to Improve Throughput on ConnectX-7 (1x400G)
with DPDK v23.07

 

Hi Ernesto,

As an additional observation, I tested the setup using iperf3 and was able
to achieve throughput of approximately 92~95 Gbps. However, when running the
DPDK-Testpmd application, the throughput does not reach similar values.

Thanks and regards,
Ayshathul Thuhara

 

  _____  

From: Ayshathul Thuhara <ayshathul.thuhara@iwave-global.com
<mailto:ayshathul.thuhara@iwave-global.com> >
Sent: 13 February 2026 14:24
To: Ernesto Ruffini <eruffini@outsys.org <mailto:eruffini@outsys.org> >
Cc: 'dev' <dev@dpdk.org <mailto:dev@dpdk.org> >; yasu1976@gmail.com
<mailto:yasu1976@gmail.com>  <yasu1976@gmail.com <mailto:yasu1976@gmail.com>
>; users@dpdk.org <mailto:users@dpdk.org>  <users@dpdk.org
<mailto:users@dpdk.org> >
Subject: Re: Guidance Required to Improve Throughput on ConnectX-7 (1x400G)
with DPDK v23.07 

 

Hi Ernesto, 

 

Thankyou for the suggestion.

I will look into
"https://fast.dpdk.org/doc/perf/DPDK_23_11_NVIDIA_NIC_performance_report.pdf
" , but here the tests uses external device for traffic generator.

 

Yes, with ethtool it shows 400Gbps.

The module can support upto 100Gbps. But it should have increased more than
10Gbps. I will disable AN and check once.

 

Thanks and regards, 

Ayshathul Thuhara

 

  _____  

From: Ernesto Ruffini <eruffini@outsys.org <mailto:eruffini@outsys.org> >
Sent: 13 February 2026 14:19
To: Ayshathul Thuhara <ayshathul.thuhara@iwave-global.com
<mailto:ayshathul.thuhara@iwave-global.com> >; users@dpdk.org
<mailto:users@dpdk.org>  <users@dpdk.org <mailto:users@dpdk.org> >
Cc: 'dev' <dev@dpdk.org <mailto:dev@dpdk.org> >; yasu1976@gmail.com
<mailto:yasu1976@gmail.com>  <yasu1976@gmail.com <mailto:yasu1976@gmail.com>
>
Subject: RE: Guidance Required to Improve Throughput on ConnectX-7 (1x400G)
with DPDK v23.07 

 

Hi Ayshathul,

I think the best is to follow
https://fast.dpdk.org/doc/perf/DPDK_23_11_NVIDIA_NIC_performance_report.pdf

Did you check with ethtool or similar the actual connection speed? ethtool
linux-nic-name | grep Speed

Maybe your module limits it to 10Gbps. In case try to disable
auto-negotiation

 

Regards,

               Ernesto

 

From: Ayshathul Thuhara <ayshathul.thuhara@iwave-global.com
<mailto:ayshathul.thuhara@iwave-global.com> > 
Sent: Friday, February 13, 2026 07:36
To: users@dpdk.org <mailto:users@dpdk.org> 
Cc: dev <dev@dpdk.org <mailto:dev@dpdk.org> >; yasu1976@gmail.com
<mailto:yasu1976@gmail.com> 
Subject: Guidance Required to Improve Throughput on ConnectX-7 (1x400G) with
DPDK v23.07

 

Dear Team,

I hope this message finds you well.

I am currently performing testing on ConnectX-7 with 1x400G hardware using
DPDK v23.07. However, I am unable to achieve data transfer rates beyond 10
Gbps. While reviewing the DPDK mailing list, I came across your previous
responses in the related discussion
(https://mails.dpdk.org/archives/users/2024-July/007767.html), where you
mentioned achieving throughput of around 250 Gbps and above.

Could you kindly advise on the configuration changes or optimizations
required to improve throughput in this setup?

For reference, I am using an OSFP loopback module for testing, where the
same port is used for both transmission and reception (single-port loopback
setup).

CX-7 Firmware: 28.45.1200

CPU : Intel(R) Core(TM) i7-14700

Command used : sudo ./build/app/dpdk-testpmd -a 01:00.0 -- -i

I would greatly appreciate your guidance.

Thank you in advance, and I look forward to your response.

Thanks and regards,
Ayshathul Thuhara

 


[-- Attachment #2: Type: text/html, Size: 14007 bytes --]

      reply	other threads:[~2026-02-13 10:09 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-02-13  6:35 Guidance Required to Improve Throughput on ConnectX-7 (1x400G) with DPDK v23.07 Ayshathul Thuhara
2026-02-13  8:49 ` Ernesto Ruffini
2026-02-13  8:54   ` Ayshathul Thuhara
2026-02-13  9:04     ` Ayshathul Thuhara
2026-02-13 10:09       ` Ernesto Ruffini [this message]

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='03ad01dc9cd0$cd2cfb00$6786f100$@outsys.org' \
    --to=eruffini@outsys.org \
    --cc=ayshathul.thuhara@iwave-global.com \
    --cc=dev@dpdk.org \
    --cc=users@dpdk.org \
    --cc=yasu1976@gmail.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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.