DPDK usage discussions
 help / color / mirror / Atom feed
From: "Talukdar, Biju" <Biju_Talukdar@student.uml.edu>
To: "Singh, Jasvinder" <jasvinder.singh@intel.com>, users <users@dpdk.org>
Subject: Re: [dpdk-users] Need help with dpdk ip pipelining
Date: Tue, 5 Apr 2016 19:11:43 +0000	[thread overview]
Message-ID: <BLUPR02MB439785EACCEEF23A8B02CD9C89E0@BLUPR02MB439.namprd02.prod.outlook.com> (raw)
In-Reply-To: <54CBAA185211B4429112C315DA58FF6DDEBFFC@IRSMSX103.ger.corp.intel.com>

Hi Jasvinder,

Thank you very much for the quick reply. I have gone through load balancing application.
One question I have is, the application output is given to the  NIC ports (for e.g. NIC  TX0, NIC TX1, NIC TX2 etc.). But my aim is to steer the output of the application to a VM. Could you please tell me how can I achieve this. Or as a matter of fact is there any generic way to hand the output of the application  to  any other usersapce process.

Thanks a ton



________________________________________
From: Singh, Jasvinder <jasvinder.singh@intel.com>
Sent: Tuesday, April 5, 2016 5:04 AM
To: Talukdar, Biju; users
Subject: RE: Need help with dpdk ip pipelining

HI Biju,

> -----Original Message-----
> From: users [mailto:users-bounces@dpdk.org] On Behalf Of Talukdar, Biju
> Sent: Monday, April 4, 2016 11:33 PM
> To: users <users@dpdk.org>
> Cc: Talukdar, Biju <Biju_Talukdar@student.uml.edu>
> Subject: [dpdk-users] Need help with dpdk ip pipelining
>
> Hi all,
>
>
> I have two questions. Could someone in the community please help. I really
> look forward to some guidance in these direction.
>
>
>     First,I have a application which implements a measurement sketch. I want
> to accelerate it with using dpdk ip pipelining application. But as I read through
> the guide in dpdk.org, I realized that it has hard pipelines for say  Rx-> flow
> classification - > ip routing -> traffic manager -> Tx. Could anyone please
> explain me how could I integrate my application in one of the pipeline. Is it
> even possible. Please help.

What do you mean by hard pipelines?  IP Pipeline application consists of reusable modules (pipeline blocks) which implement data plane functionalities such as flow classification,  routing, firewall etc. These modules are implemented using dpdk packet framework libraries (librte_port/table, librte_pipeline). In configuration file, depending upon the application, we specify various parameters of the pipeline modules such as ports (in/out), tables, core affinity, and their connectivity with each other as you can see . You can pick any pipeline module if it exists to build your application or you can implement any specific pipeline in a way existing pipeline modules have been implemented.
Please see sample configuration files such as l2fwd.cfg, l3fwd.cfg (dpdk/examples/ip_pipeline/config) to get more idea on ip pipeline application. Some other configuration files illustrate the use of various pipeline modules towards building edge router upstream /downstream packet processing workload.

>     Second, I want to use a dpdk application which can schedule packets to
> different VMs depending upon the src ip (or some logic, which is
> programmable).Is there any dpdk application available which I can use and
> modify in some way to use for the purpose. I have used OVS-DPDK for this
> purpose before. But here I  am looking for a lightweight dpdk application.
> Another condition need to be made is once I schedule  and direct a packet to
> a VM, it should be steered through usersapce only. Any pointers in this
> direction will be very helpful.

Please see load balancer sample application if it suits to your scenario.

> keep hacking
>
> Biju

  reply	other threads:[~2016-04-05 19:11 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-04-04 22:33 Talukdar, Biju
2016-04-05  9:04 ` Singh, Jasvinder
2016-04-05 19:11   ` Talukdar, Biju [this message]
2016-04-06 16:00     ` Singh, Jasvinder

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=BLUPR02MB439785EACCEEF23A8B02CD9C89E0@BLUPR02MB439.namprd02.prod.outlook.com \
    --to=biju_talukdar@student.uml.edu \
    --cc=jasvinder.singh@intel.com \
    --cc=users@dpdk.org \
    /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).