[dpdk-dev] Does anybody know OpenDataPlane

Kury Nicolas nicolas.kury at master.hes-so.ch
Thu Dec 3 10:46:46 CET 2015


Hi

Thank you everybody for your answers!

@Jim
I choose ODP as project because it was something new and I wanted to know more about the subject.  

I have read many documents about OpenDataPlane, I have tried it, implemented a small application, made some measures of throughput, etc. I know more or less what ODP is. One of my main objectives is to see how much overhead ODP adds compared to the native SDK (DPDK, USDPAA ...). Then, to try OpenDataPlane on 2 different platforms and see if my application's portability is good (same code, or few changes). I have also to see if OpenDataPlane is a "good solution" to easily port Open vSwitch on different platform. 

I have asked folks from Linaro and I have their point of view. I think it is also important to have your point of view. It is the reason why I asked you.

> Across these two spectrums and comparing ODP to DPDK is a fundamental philosophy of how much software should be open sourced vs left up to the semiconductor vendor [...]
I haven't thought about this point, interessting.

> The ARM community (ARM, Nokia, ENEA) are starting a new project called Open Fast Path http://openfastpath.org/. [...]
Yes I have seen this project. I have also seen that DPDK is also in discussion to developp a networking stack.

@Mike
> you are just beginning your research or do not understand how this fits into current telco NFV/SDN efforts.
It's difficult to get new information about OpenDataPlane on Internet. There is only a website with documentation and their philosophy. Yes, I didn't know how ODP will fit with NFV/SDN. I will focus my research in this domain, thank you for the idea.

> Maybe you can change your thesis to "Current Open Source Dataplane Methods": and do a comparison between the two.  However if you just look at the sales documentation then you may not understand the real difference.
By "DataPlane methods", do you mean virtual switches (VEB, VEPA), PCI passthrough, etc. ?

> One originated somewhat more from the embedded orientation and one originated somewhat more from the server orientation [...]
OpenDataPlane provides portability, don't you think it would be usefull to move from an orientation to another (from server to embedded or vise versa) ? Or it's something it never happens, a company never change its orientation ?
What you think is : if my orientation is server, I should use DPDK. But if my orentiation is embedded, I schould use ODP because there are lots of different kind of System on Chip and in this case, ODP would be usefull for portability. That's what you think ?

Thank you again!
Nicolas


________________________________________
De : Polehn, Mike A <mike.a.polehn at intel.com>
Envoyé : mercredi 2 décembre 2015 17:49
À : Polehn, Mike A; Kury Nicolas; dev at dpdk.org
Objet : RE: Does anybody know OpenDataPlane

A hint of the fundamental difference:
One originated somewhat more from the embedded orientation and one originated somewhat more from the server orientation. Both efforts are driving each towards the other and have overlap.

Mike

________________________________________
De : Stephen Hemminger <stephen at networkplumber.org>
Envoyé : mercredi 2 décembre 2015 17:45
À : Kury Nicolas
Objet : Re: [dpdk-dev] Does anybody know OpenDataPlane

OpenDataPlane is actually a misnomer, it is really about providing
an open (for anyone to use) API for closed source hardware platforms.


________________________________________
De : Polehn, Mike A <mike.a.polehn at intel.com>
Envoyé : mercredi 2 décembre 2015 17:32
À : Kury Nicolas; dev at dpdk.org
Objet : RE: Does anybody know OpenDataPlane

I don't think you have researched this enough.
Asking this questions shows that you are just beginning your research or do not understand how this fits into current telco NFV/SDN efforts.

Why does this exist: "OpenDataPlane using DPDK for Intel NIC", listed below? Why would competing technologies use the competition technology to solve a problem?

Maybe you can change your thesis to "Current Open Source Dataplane Methods": and do a comparison between the two.  However if you just look at the sales documentation then you may not understand the real difference.

Mike

________________________________________
De : St Leger, Jim <jim.st.leger at intel.com>
Envoyé : mercredi 2 décembre 2015 16:18
À : Kury Nicolas
Objet : RE: Does anybody know OpenDataPlane

Kury:

Congrats on selecting your masters' thesis.  What caused you to choose ODP as the subject?

On the dev at dpdk.org list are many folks from the ARM and Linaro community.  If you don't already know about Linaro and the Linaro Networking Group (LNG) you might do some initial research there.  I am sure they will respond with their views on why ODP was created the value they believe it brings to the networking world.

The ODP philosophy is fundamentally different from DPDK.  It believes in establishing a higher level common API that allows portability of applications across theoretically many architectures and many platforms underneath.  If you consider the ARM licensee model it is something necessary to abstract customers from one ARM solution not being compatible with another ARM solution (which happens today across MIPS, ARM, and I'd guess other solutions.)

>From here the semiconductor vendor would then provide an underlying SDK and/or shim layer to go from their own optimized API design (everyone has one) to the ODP API.  As with all things in software, shim layers always introduce some level of performance penalty.

Across these two spectrums and comparing ODP to DPDK is a fundamental philosophy of how much software should be open sourced vs left up to the semiconductor vendor.  The DPDK model is to do as much as possible in then open source community, all the way down to the low level polling mode drivers (PMDs.)  The ODP model leaves everything below the ODP API alone, letting the semiconductor vendor decide what to do.  In most cases they opt for a proprietary SDK, often also sold as a commercial software solution.  This is clearly in strong contrast to DPDK which is based on open and free.

One other thing to look at: The ARM community (ARM, Nokia, ENEA) are starting a new project called Open Fast Path http://openfastpath.org/. In many ways this is in contrast with the ODP model is that it aims to develop a userspace TCP/IP stack that works on top of ODP.  I personally find this a bit ironic in that the ODP concept would historically say that work of this type is NOT part of the open source community (surely not part of ODP) and thus left for the semiconductor OEMs to figure out and provide.  Maybe this is a compromise position? I don't know...

You will have to digest all of the differing views out there.  Consider also market traction.  DPDK today has a very strong, robust community with considerable adoption and uptake across the equipment and service provider customer base.  Anyone can join the project for $0 and get engaged.  Linaro projects require membership, which I believe to be quite expensive (I think you can find the data, but I believe it's $100-500k.)  I do, however, think it's possible to contribute to a Linaro product such as ODP w/o being a member.  The Linaro project and ODP has a strong list of participant names.  But the project is not nearly as mature as DPDK is today.  We'll see what the future brings.

I hope this helps some.  You can balance it against views you're likely to get from folks who are more directly involved in LNG and ODP.

Cheers,
Jim


-----Original Message-----
From: dev [mailto:dev-bounces at dpdk.org] On Behalf Of Kury Nicolas
Sent: Wednesday, December 2, 2015 7:22 AM
To: dev at dpdk.org
Subject: [dpdk-dev] Does anybody know OpenDataPlane

Hi!

Does anybody know OpenDataPlane ?  http://www.opendataplane.org/ It is a framework designed to enable software portability between networking SoCs, regardless of the underlying instruction set architecture. There are several implementations.

  *   OpenDataPlane using DPDK for Intel NIC
  *   OpenDataPlane using DPAA for Freescale platforms (QorIQ)
  *   OpenDataPlane using MCSDK for Texas Insturments platforms (KeyStone II)
  *   etc.

When a developer wants to port his application, he just needs to recompile it with the implementation of OpenDataPlane related to the new platform.

I'm doing my Master's Thesis on OpenDataPlane  and I have some questions.

- Now that OpenDataPlane (ODP) exists, schould every developpers start a new project with ODP or are there some reasons to still use DPDK ? What do you think ?

Thank you very much

Nicolas




More information about the dev mailing list