[Opensdwmn] Fwd: Re: Introduction to open source MPLS solution

Tue Apr 2 08:50:29 CEST 2019

Gracias por la revisión Pedro, es muy facilitadora.

Mis sensaciones, después de pensar, de charlar ayer del tema con Nacho y de leer tu mensaje: en OC han avanzado bastante en los nodos de acceso, pero no hay nada relevante en backhaul, salvo la voluntad de meter bajo control al backhaul en el mismo controlador que el acceso. No creo que haya mucho de lo que partir, así que bien podría ser que la invitación a participar en OC-backhaul se acabe convirtiendo en la invitación a tirar de OC-backhaul, con más apoyos, posiblemente con dinero, pero también con más prisas.

Yo creo que es un reto muy interesante y me gustaría decir que "adelante"; mi única reserva es que yo en este momento tengo muy limitada mi dedicación temporal, y me da un poco de miedo meternos en algo cuya velocidad no pueda seguir. Además, me parece que preguntar sobre eso en la conf-call no es una buena tarjeta de visita. Yo no me animaría mucho a colaborar con alguien que de partida me dice que no quiere correr mucho.

Si os parece contesto a Kashif, propongo una conf-call para la semana que viene y vemos. Algun at s no habéis dicho nada, valoraría vuestra opinión.

Muchas gracias

Javier

El 2/4/19 a las 8:24, Pedro de las Heras Quirós escribió:

Hola,

Aporto en este correo: a) referencias que pueden ayudar a entender
OpenCellular, b) preguntas/dudas que me han surgido según leía sobre
OpenCellular, c) reflexiones/sugerencias para la telco.

a) Referencias sobre OpenCellular

0) Esta es la referencia de mi otro correo-e, la pongo para que estén
aquí juntas:
https://cacm.acm.org/magazines/2018/8/229756-designing-sustainable-rural-infrastructure-through-the-lens-of-opencellular/fulltext

1) Esto es lo mejor que he encontrado para entender técnicamente la
BTS OpenCellular:
Charla de Emily McMilin + Kashif Ali: OpenCellular: Open Source
Wireless Access Platform
Vídeo
https://media.ccc.de/v/osmocon17-4013-opencellular_open_source_wireless_access_platform
Transparencias
https://osmocom.org/attachments/2623/opencellular.pdf

En transpas 12,13,14 aparece la arquitectura SW/Firmware, si bien hay
que tener en cuenta que en su correo Kashif dice que están
redefiniendo el stack actualmente.

La charla se centra en el HW y Emily al final en la parte radio. La
parte SW la tratan muy por encima. Me hace pensar que tienen muy poco
pensado de backhaul. Más al respecto después.

2.a) Este artículo explica el Community Cellular Manager: es el SW
para operar la BTS, que usan para gestionar redes que usan Open
Cellular. Aparece en la transparencia 12 de la anterior referencia
(módulo superior de la pila). Es sw con python/django.
https://www.usenix.org/conference/nsdi19/presentation/hasan
2.b) Vídeo sobre Community Cellular Manager:
https://media.ccc.de/v/osmocon17-4014-community_cellular_manager

3) Usan como stack GSM/LTE https://osmocom.org/ Lo último que he visto
de esta gente es que con SDR ponen GSM sobre WiFi:
https://abopen.com/news/osmocombb-sdr-phy-project-achieves-gsm-on-any-frequency-milestone/

4) Sobre osmocom, esta charla de 2018 es trabajo relacionado: Redes
GSM en Brasil/Mexico, implementadas con osmocon (pila gsm que usa
opencellular), con backhaul WiFi. No es técnica, la he visto a doble
velocidad, no aporta demasiado para mí técnicamente, salvo la
evidencia de que osmocom funciona y se está desplegando:
https://media.ccc.de/v/osmocon2018-67-community-cellular-network-implementations-in-latin-america

4) Sobre OC-backhaul/OC-controller mencionados por Kashif en su
correo-e: me da la impresión de que los proyectos
OC-backhaul/OC-controller que menciona Kashif en sus correos-e son
demasiado incipientes pues no encuentro mucho al respecto en la red ni
en el github de OpenCellular. Decía antes que no aparece mencionado en
su charla ni artículo (ver referencias anteriores 0 y 1). Sigo
buscando sobre OC-backhaul/OC-controller.

5) SW de OpenCellular: https://github.com/Telecominfraproject/OpenCellular

------------------------------------------------
b) Preguntas que me han surgido:

Me da la impresión, viendo las descripciones del hw libre de
OpenCellular y leyendo las referencias en a), que el target de
OpenCellular es 2G: que las comunidades rurales puedan usar sus
móviles para hablar y sms. En el futuro LTE. Su HW no ofrece ifaces
WiFi. Entiendo, pero es más una pregunta que una afirmación, que en
esto difieren del target de Tucan3G /OpenSDWMN.  ¿Estoy en lo cierto?
Cualquier aclaracion respecto a la duda que planteo me/nos será de
ayuda para entender la visión e intersección de OpenCellular y
OpenSDWMN, y por tanto espero que ayude a decidir sobre el interés de
hacer cosas en el paraguas OpenCellular.

En relación a esto, el stack osmo para GSM que usa OpenCellular parece
algo realmente interesante (aparte de OpenCellular, ver referencia 3
en sección anterior). Lo vi por primera vez en una demo del tadhack
que co-organizamos hace años. Pero esto ha avanzado muchísimo, con un
muchos proyectos diferentes.

c) En el correo-e de Kashif dice: "We have x86 (E3845) and Arm (new
SoC coming from Marvell) based open-source hardware for computing,
which run SDR stack and NoC, and will be able to support the MPLS.
What is your hardware requirement?"

Entiendo que este hw del que habla es el de la BTS descrito en la
referencia 1) más arriba. Si es así, a la pregunta que hace entiendo
que habría que responder que la configuración de HW que tenemos en
mente es algo más parecido a LibreRouter, es decir, interfaz/es WiFi
2.4GHz y al menos 2 WiFi 5G.

Imagino que lo que ellos harán para conectar al backhaul su BTS es
conectar hw adicional via Ethernet, y para eso es para lo que pide
necesidades hw. Pero es mi interpretación.

También le preguntaría por todo lo que pueda contar de
OC-backhaul/OC-controller, arquitectura que están redefiniendo, todo
sobre el controlador, papel de SDN en su arquitectura.

Finalmente, le preguntaría claro por el tipo de vinculación que pide y
el tipo de financiación del que habla.

Salud,
Pedro

Salud,
--
Pedro de las Heras Quirós
Departamento de Teoría de la Señal y Comunicaciones y Sistemas
Telemáticos y Computación
Universidad Rey Juan Carlos - Campus de Fuenlabrada (Madrid), España
pedro.delasheras at urjc.es<mailto:pedro.delasheras at urjc.es> | www.urjc.es<http://www.urjc.es> | @ETSIT_URJC
On Mon, Apr 1, 2019 at 1:42 PM Pedro de las Heras Quirós
<pedro.delasheras at urjc.es><mailto:pedro.delasheras at urjc.es> wrote:

Pedro de las Heras Quirós <pheras at gmail.com><mailto:pheras at gmail.com>
Hola,

Es difícil decidir sin saber algo más. Lo que parece claro a juzgar
por lo que yo había oído estos años de Telecom Infra y por lo que
estoy leyendo de Open Cellular, es que su compromiso con tecnología
hw/sw abierta es total. El propio correo-e habla de las presiones a
WiBack para que liberen.

Respecto a si su plataforma hw/sw es más o menos interesante, estoy
mirando qué información hay por ahí de OC-Backhaul. No encuentro
mucho, debe ser muy reciente.

Voy mandando lo que encuentro:
https://cacm.acm.org/magazines/2018/8/229756-designing-sustainable-rural-infrastructure-through-the-lens-of-opencellular/fulltext

Puedo estar en la telco si soy de ayuda.

Salud,
Pedro

--
Pedro de las Heras Quirós
Departamento de Teoría de la Señal y Comunicaciones y Sistemas
Telemáticos y Computación
Universidad Rey Juan Carlos - Campus de Fuenlabrada (Madrid), España
pedro.delasheras at urjc.es<mailto:pedro.delasheras at urjc.es> | www.urjc.es<http://www.urjc.es> | @ETSIT_URJC
On Mon, Apr 1, 2019 at 11:15 AM Javier Simó Reigadas
<javier.simo at urjc.es><mailto:javier.simo at urjc.es> wrote:

Estimados

El otro día os comenté que Carlos Rey me había puesto en contacto con Kashif Ali, del proyecto Open Cellular de Facebook, pero que de momento no había contestado. Pues ya ha contestado, y la pelota vuelve a estar en nuestro tejado. Si yo entiendo bien su correo, nos propone:

1.- Si nos interesa, "integrarnos" en el subproyecto OC-Backhaul, que iría controlado por el mismo controlador que la red de acceso radio. Supondría condicionarnos al "paraguas" de OpenCellular, compartir código, ...

2.- y en ese caso podría haber "alguna financiación".

3.- Tener una conferencia telefónica para discutirlo.

Yo, antes de tener la conferencia telefónica querría contar con vuestro feedback. Entrar al paraguas de OpenCellular es algo que puede ser muy potente, al tiempo que muy demandante y algo restrictivo. Estaría bien decidir bajo qué criterios negociamos (p.e., estamos dispuestos a cualquier cosa siempre que sea open-source, o queremos además que sea sí o sí SDN con oopenWRT y openvSwitch?). Una vez claros en esto, no sé quién es la mejor persona para mantener una conferencia telefónica, yo solo desde luego no porque siempre he tenido cierto hándicap para entender el inglés de los nativos por teléfono.

Saludos

-------- Mensaje reenviado --------
Asunto: Re: Introduction to open source MPLS solution
Fecha: Sun, 31 Mar 2019 16:23:01 +0000
De: Kashif Ali <kashi at fb.com><mailto:kashi at fb.com>
Para: javier.simoATurjc.es <javier.simo at urjc.es><mailto:javier.simo at urjc.es>, Carlos <carlos at apc.org><mailto:carlos at apc.org>

Hello Javier,

Thanks Carlos, for the introduction and my apology for late response, as I wrote the draft and thought I have sent already!

Javier, great to meet you and thanks for providing full context. I did my post-doc work with Eric Brewer at UC Berkeley, so I know bunch about the WiLD project. Great work.

On my end, I have started OC-Backhaul project (as part of OpenCellular) and have been trying to convince WiBack to open-source their stack but apparently FiT management doesn’t want. Their model is to license their stack which doesn’t fit with OpenCellular: open-source and royalty-free. OC deployment model is full rural site solution, including spectrum management, power, cellular (2G/LTE), etc and ofcourse backhaul. For some of these sub-projects, we are far into development  (and even production deployments) and others early idea/design stage.

OC-backhaul project will be mix of SAT and MPLS-based mesh, integrated with OC controller (we are in middle of re-arch the whole OC stack and would want to have single controller managing the network). We have x86 (E3845) and Arm (new SoC coming from Marvell) based open-source hardware for computing, which run SDR stack and NoC, and will be able to support the MPLS. What is your hardware requirement?

If you open to work under the OC umbrella, contribute code and integrate with the OC architecture, so we won’t end up replicating or use resources to have two open-source implementations for same thing, I think there is great synergy and ofcourse, availability of some funding.

Maybe we should schedule a conf call sometime this week to discuss?

Looking forward to chatting with you.

Thanks

Kashif

From: Javier Simó Reigadas <javier.simo at urjc.es><mailto:javier.simo at urjc.es>
Date: Thursday, March 21, 2019 at 6:55 AM
To: Carlos <carlos at apc.org><mailto:carlos at apc.org>, Kashif Ali <kashi at fb.com><mailto:kashi at fb.com>
Subject: Re: Introduction to open source MPLS solution

Dear Kashif,

As Carlos says, we have been working on rural backhaul networks for many years now. Back in 2003 we started to consider multi-hop wireless rural networks based on WiLD (WiFi over Long Distances), initially with telemedicine purposes only. Our first networks of that type connected remote rural health facilities with reference hospitals in Peru, Colombia, Ecuador and Cuba. After many years accumulating experience with technology, logistics, management, etc. we saw that it was not reasonable that our rural networks, always in regions where no other telecommunications infrastructures were present, could be exploited for telecommunications services other than telemedicine. So we decided that it was time to collaborate with operators and see how those rural transport networks could be both telemedicine networks and the backhaul for rural base stations for general-purpose mobile communications.

That is why we started the TUCAN3G project in 2013. In 2016 we had demonstrated how rural networks based on multi-hop low-cost wireless networks could act as carrier-class backhaul solutions for rural femtocells, and the overall solution would allow operators to extend services to very small and remote villages with a sustainable business model, specially if infrastructures were shared with communities or public administrations. Our backhaul networks required initial planning and had tree topology; we never worked on dynamic routing or self-configuring topologies. On the other hand, we handled very carefully QoS issues, differentiating signalling, voice traffic and data and providing guarantees for end-to-end QoS as recommended by ITU G.1010 for each service. Moreover, we designed distributed optimization algorithms that could reallocate network resources among the different traffic classes and source nodes dynamically as the traffic load and the link conditions change, always ensuring priority and quality at least for voice and signaling traffic. We demonstrated our results in two real testbeds, one in the Napo river managing traffic QoS with a DiffServ approach, and the other one in the Paranapura river based on MPLS (both in the Amazon rainforest in Peru). Real users received services through those networks, though Telefonica del Perú discontinued the service at the end of the project, at least in the Paranapura river.

After completion of the TUCAN3G project, we contacted the Fraunhofer institute because we wanted to know more about the WiBACK technology. We signed an agreement with them in which they accepted to provide us with their software for integration with our own embedded WiFi routers in laboratory for free, and we accepted not to distribute the software or to publish any results or observations without their supervision. For one year we tested the WiBACK solution with our hardware, which was somehow problematic because there were bugs that had been corrected in the past for their official ARM hardware but not for x86 platforms as ours. Their support was incredibly good for the whole process, with good responses to all our questions always in less than 24H. We found out that WiBACK was very interesting, having a very good architectural base, really self-configuring in terms of topology and routing. There is a good support for traffic classification and mapping into MPLS classes, with central management in the coordinator node. The whole thing is very well conceived. Our observations were not that positive in terms of real carrier-class QoS support, as we did not find a perfect matching between the QoS and the capacity assigned for each flow and what actually happened, but that could be because of our measurements. I mean, the evaluation was run by a student with very little time for the project, and he could perfectly have made mistakes. So I cannot be sure about the real nature of the "issues" we found. However, we saw that there was little flexibility in the way resources were allocated for the different LSPs, and how priorities were applied; that did not seem to be as flexible as topology self-configuration.

Then we got some funding for a pre-doc full-time grant, and one person started working with us three weeks ago. We met (our team are 6 professors with partial dedication to this project, and a full-time pre-doc student), we discussed about what to do for the next 12 months, and we decided to leave WiBACK aside for the moment and focus on how to implement an open-source SDN version of what we did for TUCAN3G, but with improvements in the self-configuring part that we have seen so good in WiBACK. Here we are now, starting this project.

Although we are in the first steps of this new project, we would be working in parallel with the functional definition, software development in a simulation platform (mininet + openvSwitch) and identifying an appropriate hardware for the moment we want to go into real testbeds. We are also trying to get some libreRouter devices to see if, adding a openvSwitch to them, we can chose them as our hardware, as they meet most of our requirements.

In parallel, one of our researchers in working with two students in implementing a virtualized mobile network in the lab with OpenAirInterface and B210 Ettus SDR devices. We were also interested at some point last year in OpenCellular, but we were told (may be wrongly?) that it was not possible to use it for LTE or beyond by that time.

All I could add is that we have experience and ideas, but scarce resources that limit the speed and scope of the objectives we can have for the short term.

I hope this is not "too" long as a detailed presentation. Please do not hesitate to make any questions about what we are/know/can do and what we are not/don't know/cannot do. Best wishes,

Javier

El 21/3/19 a las 10:56, Carlos escribió:

Hey Kashif, I wanted to follow up on the conversation we had yesterday

and introduce you to Javier, cc'ed.

Javier, besides having been formally my PhD supervisor, is one of the

person I respect and admire the most in this space, and a personal

friend. He has been involved in rural backhaul solutions for many years

[1], [2]. Over the past few years, Javier has been involved in Tucan3G

[3], where Mayutel (FB's partner) is also involved, focusing on

developing an open source MPLS-type solution to prioritize the traffic

in the backhaul, but I will let him to tell you more about it. His

research group has also been analyzing WiBack's performance, and were

evaluating contributing to the open API that Matthias was considering,

but a couple of weeks ago they decided to work on a full open source

solution providing an automatic configuration via SDN to the MPLS

priorities developed in Tucan3g. Additionally, Javier was telling me

yesterday that his research team is also working with some OAI

solutions, to integrate access and backhaul. So, all in all it looks to

me that you two make a perfect match. I leave it to you to continue this

conversation!

All the best,

carlos

[1] https://scholar.google.es/citations?user=zoKJHR4AAAAJ&hl=es

[2] https://www.researchgate.net/profile/Javier_Simo

[3] https://www.youtube.com/watch?v=i-CLahBb_4Q

--

Francisco Javier Simó Reigadas

Profesor, Director de la ETSIT

Dept. Teoría de la Señal y Comunic. y Sist. Telemáticos y Computación

Universidad Rey Juan Carlos

Campus de Fuenlabrada (Madrid), España

Teléfono: 91 488 81 67

javier.simo at urjc.es<mailto:javier.simo at urjc.es> | www.urjc.es<http://www.urjc.es> | @URJC |

_______________________________________________
Opensdwmn mailing list
Opensdwmn at gsyc.urjc.es<mailto:Opensdwmn at gsyc.urjc.es>
https://gsyc.urjc.es/cgi-bin/mailman/listinfo/opensdwmn

_______________________________________________
Opensdwmn mailing list
Opensdwmn at gsyc.urjc.es<mailto:Opensdwmn at gsyc.urjc.es>
https://gsyc.urjc.es/cgi-bin/mailman/listinfo/opensdwmn

--

[cid:part1.0BB35966.3F1DEB14 at urjc.es]

Francisco Javier Simó Reigadas

Profesor, Director de la ETSIT

Dept. Teoría de la Señal y Comunic. y Sist. Telemáticos y Computación

Universidad Rey Juan Carlos

Campus de Fuenlabrada (Madrid), España

Teléfono: 91 488 81 67

javier.simo at urjc.es<mailto:javier.simo at urjc.es> | www.urjc.es<http://www.urjc.es> | @URJC |

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://gsyc.urjc.es/pipermail/opensdwmn/attachments/20190402/5468e34d/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Imagen 1
Type: image/png
Size: 5104 bytes
Desc: Imagen 1
URL: <http://gsyc.urjc.es/pipermail/opensdwmn/attachments/20190402/5468e34d/attachment-0001.png>