On 03/11/2014 05:52 PM, Ricardo de Freitas Gesuatto wrote:

Hello Maxim,

Thank you for the feedback. I have started writing a more formal application, following the Template from the wiki. After some thought, I have two possible projects on top of OpenDataPlane:

1. Lightweight IP Stack on OpenDataPlane. This is the idea originally suggested in the wiki, albeit in more detail regarding approach. An IP stack could provide more visibility to the project and lead to more applications ported to ODP in the future.

2. Porting Libfluid to OpenDataPlane. Libfluid was recently endorsed by the ONF as their "OpenFlow driver", and can be used for building OpenFlow-enabled applications (mostly OF controllers, but a sample software switch already exists).

that is interesting idea but code to be included to odp has to fit odp code style: C not C++, BSD license and etc. Some people might already have some work done to support open flow. So it's better to subscribe on our mailing list lng-odp@linaro.org and send this idea proposal there. Which I recommend you to do it.

We will have OpenVSwitch working on top on ODP. And because of libfluid works with ovs it will be there. However there might be performance benefit using openflow on odp itself. At this time we do not have flow tables support there. Just because it's new project. But flow tables will be there very soon. Because of openflow itself is simple I would implement switch side inside the ODP and for controller side used existence solutions.

The application I'm writing focuses on idea #1 as it likely benefits Linaro the most. Idea #2 is interesting as a research in network performance, but due to project scope it may be be unsuitable for GSOC. However, depending on the mentor's decision/interest I may still write it down as an alternative application.

Also, is there a preferred way to share the application before submitting?

Google code is new for all of us. But you can send application to us for review.

Thanks,

Ricardo de Freitas Gesuatto

Odp info can be found here:

http://www.opendataplane.org/ https://bugs.launchpad.net/linaro-odp https://git.linaro.org/lng/odp.git http://www.opendataplane.org/wp-content/uploads/2014/01/ODPIntroductionandOv... API: build Doxygen docs with make docs

Thank you, Maxim.

On Sat, 8 March 2014 23:35, Maxim Uvarov maxim.uvarov@linaro.org wrote: Hello Ricardo,

recieved email. I'm traveling at this time and will contact you later in a week.

We have remote access to different hardware in LAVA framework. You can find some information about it here: https://wiki.linaro.org

However for iwip+odp development raw sockets can be used and after that check speed on accelerated SoCs.

At this time we also interested in some applications which can be ported to odp.

Thanks, Maxim.

On 03/07/2014 05:41 AM, Ricardo de Freitas Gesuatto wrote:

...

Hello,

I'm currently a Computer Science student at the Federal University of Sao Carlos, Brazil. I have some background in operating systems and (mainly) networking technologies, from academic networking projects to OpenFlow research & development. I am interested in participating in this Google Summer of Code, as one of the project ideas on the Linaro Wiki ignited my curiosity...

The OpenDataPlane project is really interesting, as it can provide high-performance, hardware-offloaded throughput for networking applications. From an outsider point of view, ODP seems like the answer (from the ARM universe) to Intel's DPDK, but with the advantage of being truly open and portable.

An outreach program such as GSOC allows students to leave their comfort zone, to learn something new and contribute to a project. Albeit I have no prior OpenDataPlane knowledge, I have been reading the available documentation for ODP since the mentoring organizations list was announced.

In the long run, I expect to study if it's feasible to use OpenDataPlane as a lower level API for implementing OpenFlow components, and eventually port and run OpenFlow on top of ODP. However, the 12 weeks for the GSOC is likely too short for such a project.

On the other hand, the "lightweight IP stack on top of OpenDataPlane" project suits the GSOC very well due to well-defined and progressive milestones. Developing this project in an iterative fashion would reduce the risks associated with the GSOC schedule. Plus, having an IP stack on top of ODP could potentially increase the project's visibility :)

Despite all my interest, there's still a catch... I don't have any ODP-enabled SoC at university. I have read on the website about the Linux-based environment and API. Would it suffice for the GSOC project? It's an important question, as this may affect the performance evaluation proposal at the end of the project.

I think that's it for now. Please post any questions / answers / critics / suggestions, as I would like to know if I'm on the right way,

Thanks,

Ricardo Gesuatto

---

linaro-gsoc mailing list linaro-gsoc@lists.linaro.org http://lists.linaro.org/mailman/listinfo/linaro-gsoc