Re: auto ip address assignment
Pablo Osuna <posuna <at> ehas.org>
2006-04-04 11:40:44 GMT
Hi:
I am really happy to see this discussion in olsr-list. From my position
I am also VERY interested on this issue. Some months ago I briefly
discussed this issue with Andreas but forgot to post the stuff in the
mailinglist. But let me first introduce myself :) My name is Pablo
Osuna. I am part of EHAS (www.ehas.org), a spanish non-profit
institution (NGO), whose ultimate goal is improving public health
assistance services in rural areas of Latin American countries. We are
mainly focused in wireless technologies like Wi-Fi or VHF-HF. I am the
person in charge of developing a wireless mesh router with solar energy
and supported with an embedded Linux system. This router will be part of
a mesh IP network, aimed to provide both data and voice services. Our
first prototype is based on Soekris and pebble-linux. At the moment we
are using wireless cards running with hostap and madwifi drivers. And
now we are busy improving this scheme in different aspects:
1) migration to a low-powered architecture based on ARM or MIPS.
2) IP autoconfiguration.
3) Dynamic multi-hop routing protocol QoS aware.
4) QoS at IP layer in order to ensure a stable VoIP connection.
We have found olsrd a very good multi-hop dynamic protocol. I have to
say congratulations for such a good job, and especially for all the work
behind in the maintenance and upgrading. We are an NGO with very scarce
resources, and projects like yours help us a lot.
The final idea of our project is that nodes do automatically every thing
once they are placed in the right positions. As a node is powered on, it
discovers its neighbours, attributes itself a unique IP address, and
then establish the most appropriated routes to the rest of the network
and to the world, taking into account the quality of the links. Nodes
know about the different classes of traffic in the network, and manage
the QoS (Quality of Service) at the IP level. Additionally, each node
has the necessary elements to make possible partial voice and data
connectivity even if part of the network is provisionally not available.
In short if we are able to obtain a highly self-configurable mesh
architecture, nodes can be place at strategical positions by any
non-qualified person whose only concern will be to assure the
line-of-sight to neighbour nodes. And believe me, in developing
countries IT staff resources are very expensive and scarce.
With olsrd we solve one important requirement, the dynamic routing.
Autoconfiguration is the next issue to face. We are a bit frustrated
with this point because although there are plenty of theoretical
research on it (please see a good draft about this on
http://www.it.uc3m.es/cjbc/papers/draft-bernardos-manet-autoconf-survey-00.html)
we can not say the same on the implementation side. Some partial or
unfinished solutions we have seen until now are:
- In IPv6 there are several solutions inherent in the protocol itself to
obtain automatically an IP address in the scope of an unique multicast
domain. This local character makes them useless for the nodes of a Mesh
network whose addresses do not keep a fixed relation and therefore do
not share the same domain. It would be necessary to propagate the
network prefix (Andreas suggested once to use olsrd to carry the
information).
- In IPv4 ZEROCONF workgroup has implemented a similar solution to IPv6
one. Using this implementation (zcip in Linux) a node is automatically
set up with a 169.254/16 IPv4 address. As the case of IPv6 these
addresses are just valid for the communication with nodes in the same
physical link (or logical). Also in IPv4, although a node has a unique
48 bits MAC address, there are no means to obtain a unique 32 bits IPv4
address from it. One partial solution in order to get this uniqueness
could be the use of wireless cards from the same manufacturer (very
limiting idea :( )
- Besides PACMAN we also know this project NOA-OLSR
(http://hipercom.inria.fr/noa-olsr/) and LUNAR-ng based on AODV
(http://core.it.uu.se/AdHoc/LunarImpl) but all of them offer limited
solutions, some of them already discussed here: no QoS-aware, limited
number of interfaces per node, inestability.
- A DHCP centralized server would permit that a node could get its IP
address making use of several DHCP relays in a chain (I do not know if
this would be possible) or sending DHCP requests through routing olsrd
packets. But I do not like this idea so much, it is very depending on
the network links and it is not a decentralized idea typical of Mesh
networks.
We also agree that implementing a plugin for olsrd would be the best
idea. olsrd has achieved a remarkable maturity and stability and
nowadays it keeps being supported by an extensive community.
We will follow this issue in a very close way. Unfortunately we do not
have developers at this time to put efforts on this.
Thanks!
Pablo
El lun, 03-04-2006 a las 17:54 -0500, olsr-dev-request <at> olsr.org
escribió:
> Send olsr-dev mailing list submissions to
> olsr-dev <at> olsr.org
>
> To subscribe or unsubscribe via the World Wide Web, visit
> https://www.olsr.org/mailman/listinfo/olsr-dev
> or, via email, send a message with subject or body 'help' to
> olsr-dev-request <at> olsr.org
>
> You can reach the person managing the list at
> olsr-dev-owner <at> olsr.org
>
> When replying, please edit your Subject line so it is more specific
> than "Re: Contents of olsr-dev digest..."
>
>
> Today's Topics:
>
> 1. auto ip address assignment (Abhishek Misra)
> 2. Re: auto ip address assignment (Andreas T?nnesen)
> 3. Re: auto ip address assignment (Sven-Ola Tuecke)
> 4. Re: auto ip address assignment (Pitu)
> 5. Re: auto ip address assignment (Jens Nachtigall)
> 6. Re: auto ip address assignment (Pitu)
> 7. RE: auto ip address assignment (Dan Flett)
>
>
> ----------------------------------------------------------------------
>
> Message: 1
> Date: Mon, 3 Apr 2006 15:23:29 +0530 (IST)
> From: Abhishek Misra <abhishekm <at> cdac.in>
> Subject: [olsr-dev] auto ip address assignment
> To: <olsr-dev <at> olsr.org>
> Message-ID: <twig.1144058009.1784 <at> cdac.in>
>
>
> Hello,
>
> i was going through a ppt in docs titled 'mobile adhoc networks'
> there you have mentioned auto ip address assignment scheme , well
> actually
> its not mentioned ! OK please tell what do you mean by 'New nodes are
> assigned unused IP addresses through use of strong DAD'. please clearify
> 'strong DAD'
>
> bye
>
>
>
>
>
>
> ------------------------------
>
> Message: 2
> Date: Mon, 03 Apr 2006 12:08:43 +0200
> From: Andreas T?nnesen <andreto <at> olsr.org>
> Subject: Re: [olsr-dev] auto ip address assignment
> To: OLSR development <olsr-dev <at> olsr.org>
> Message-ID: <4430F42B.3040600 <at> olsr.org>
> Content-Type: text/plain; charset=ISO-8859-1; format=flowed
>
> Hi,
>
> The autoconfig scheme is explained in my master:
> http://www.olsr.org/docs/report_html/node176.html
> The Strong/Weak DAD terms are explained there.
>
> Let me once again state that the auto-config implementation was not
> open source and the code is not available.
>
> - Andreas
>
> Abhishek Misra wrote:
> > Hello,
> >
> > i was going through a ppt in docs titled 'mobile adhoc networks'
> > there you have mentioned auto ip address assignment scheme , well
> > actually
> > its not mentioned ! OK please tell what do you mean by 'New nodes are
> > assigned unused IP addresses through use of strong DAD'. please clearify
> > 'strong DAD'
> >
> > bye
> >
> >
> >
> >
> >
> > _______________________________________________
> > olsr-dev mailing list
> > olsr-dev <at> olsr.org
> > https://www.olsr.org/mailman/listinfo/olsr-dev
>
>
>
> ------------------------------
>
> Message: 3
> Date: Mon, 3 Apr 2006 12:32:40 +0200
> From: "Sven-Ola Tuecke" <mail2news <at> commando.de>
> Subject: Re: [olsr-dev] auto ip address assignment
> To: olsr-dev <at> olsr.org
> Message-ID: <e0qtk8$hh0$1 <at> gate.commando.de>
> Content-Type: text/plain; format=flowed; charset="iso-8859-1";
> reply-type=original
>
> Hey,
>
> funny. Everyone is fiddeling with a rather boring variant of DHCP. Why in
> hell nobody has a sense exciting stuff? What about OLSR radio!?
>
> LG Sven-Ola
>
> "Abhishek Misra" <abhishekm <at> cdac.in> schrieb im Newsbeitrag
> news:twig.1144057950.65528 <at> cdac.in...
> > Hello,
> >
> > i was going through a ppt in docs titled 'mobile adhoc networks'
> > there you have mentioned auto ip address assignment scheme , well
> > actually
> > its not mentioned ! OK please tell what do you mean by 'New nodes are
> > assigned unused IP addresses through use of strong DAD'. please clearify
> > 'strong DAD'
> >
> > bye
> >
> >
> >
> >
> >
> > _______________________________________________
> > olsr-dev mailing list
> > olsr-dev <at> olsr.org
> > https://www.olsr.org/mailman/listinfo/olsr-dev
>
>
>
>
> ------------------------------
>
> Message: 4
> Date: Mon, 3 Apr 2006 15:29:19 +0200
> From: Pitu <pitu <at> aluzina.org>
> Subject: Re: [olsr-dev] auto ip address assignment
> To: "OLSR development" <olsr-dev <at> olsr.org>
> Message-ID:
> <fb5d17400604030629x6c9f8533o35b37590552933ae <at> mail.gmail.com>
> Content-Type: text/plain; charset="iso-8859-1"
>
> Is anybody willing to develop the autoconfig plugin? Me and a colleague
> would like to implement it, but up to now, we're only looking for more
> information (we're putting all together
> here<http://www.aluzina.org/trac/wiki/MANETs>).
> We've thought that using any existing implementation could be a good idea
> (like PACMAN <http://sourceforge.net/projects/pacman-autoconf/>).
>
> Do you know any other alternative?
>
> Pitu.
>
> 2006/4/3, Sven-Ola Tuecke <mail2news <at> commando.de>:
> >
> > Hey,
> >
> > funny. Everyone is fiddeling with a rather boring variant of DHCP. Why in
> > hell nobody has a sense exciting stuff? What about OLSR radio!?
> >
> > LG Sven-Ola
> >
> > "Abhishek Misra" <abhishekm <at> cdac.in> schrieb im Newsbeitrag
> > news:twig.1144057950.65528 <at> cdac.in...
> > > Hello,
> > >
> > > i was going through a ppt in docs titled 'mobile adhoc networks'
> > > there you have mentioned auto ip address assignment scheme , well
> > > actually
> > > its not mentioned ! OK please tell what do you mean by 'New nodes are
> > > assigned unused IP addresses through use of strong DAD'. please clearify
> > > 'strong DAD'
> > >
> > > bye
> > >
> > >
> > >
> > >
> > >
> > > _______________________________________________
> > > olsr-dev mailing list
> > > olsr-dev <at> olsr.org
> > > https://www.olsr.org/mailman/listinfo/olsr-dev
> >
> >
> > _______________________________________________
> > olsr-dev mailing list
> > olsr-dev <at> olsr.org
> > https://www.olsr.org/mailman/listinfo/olsr-dev
> >
> -------------- next part --------------
> An HTML attachment was scrubbed...
> URL: http://www.olsr.org/pipermail/olsr-dev/attachments/20060403/aa01c685/attachment.htm
>
> ------------------------------
>
> Message: 5
> Date: Mon, 3 Apr 2006 17:38:27 +0200
> From: Jens Nachtigall <nachtigall <at> web.de>
> Subject: Re: [olsr-dev] auto ip address assignment
> To: OLSR development <olsr-dev <at> olsr.org>
> Message-ID: <200604031738.34197.nachtigall <at> web.de>
> Content-Type: text/plain; charset="iso-8859-15"
>
> > We've thought that using any existing implementation could be a good idea
> > (like PACMAN <http://sourceforge.net/projects/pacman-autoconf/>).
>
> I had a look at PACMAN a year ago, it is not working with the current olsrd
> release, also it only works on linux. But the ideas of pacman as described in
> the paper, are quite good, if I remember right. So if were you, i would read
> that paper and reimplement these ideas all from scratch in a olsrd plugin.
> The algos in the paper are enough to get it working, afaikr.
>
> Please also note, that nobody uses the original olsr-rfc-protocol, but rather
> the enhancements like etx. I think in the freifunk firmware, which most
> meshes use, also the MPRs are switched off, i.e. every node behaves linke an
> MPR. So the hierarchie is flat. This is because, the MPR idea did not work
> out.
>
> Good luck on implementing. Rather get started, there are already so many
> papers on meshes, but code is rare 
>
> All the best,
>
> jens
> -------------- next part --------------
> A non-text attachment was scrubbed...
> Name: not available
> Type: application/pgp-signature
> Size: 189 bytes
> Desc: not available
> Url : http://www.olsr.org/pipermail/olsr-dev/attachments/20060403/ac78da27/attachment-0001.pgp
>
> ------------------------------
>
> Message: 6
> Date: Mon, 3 Apr 2006 18:16:09 +0200
> From: Pitu <pitu <at> aluzina.org>
> Subject: Re: [olsr-dev] auto ip address assignment
> To: "OLSR development" <olsr-dev <at> olsr.org>
> Message-ID:
> <fb5d17400604030916s72f876dai76e1d57c9191dadf <at> mail.gmail.com>
> Content-Type: text/plain; charset="iso-8859-1"
>
> 2006/4/3, Jens Nachtigall <nachtigall <at> web.de>:
> >
> > > We've thought that using any existing implementation could be a good
> > idea
> > > (like PACMAN <http://sourceforge.net/projects/pacman-autoconf/>).
> >
> > I had a look at PACMAN a year ago, it is not working with the current
> > olsrd
> > release, also it only works on linux. But the ideas of pacman as described
> > in
> > the paper, are quite good, if I remember right. So if were you, i would
> > read
> > that paper and reimplement these ideas all from scratch in a olsrd plugin.
> > The algos in the paper are enough to get it working, afaikr.
>
>
> Yes, it's the path that we would follow. PACMAN works as a daemon such that
> olsrd is not aware of his existence (as far as I'm concerned). The idea was
> to turn this code into an olsrd plugin in order to work closely with olsrd.
>
> Please also note, that nobody uses the original olsr-rfc-protocol, but
> > rather
> > the enhancements like etx. I think in the freifunk firmware, which most
> > meshes use, also the MPRs are switched off, i.e. every node behaves linke
> > an
> > MPR. So the hierarchie is flat. This is because, the MPR idea did not work
> > out.
>
>
> I didn't know that... it sounds quite frustrating. MPR seemed to be a good
> idea...
>
>
> Good luck on implementing. Rather get started, there are already so many
> > papers on meshes, but code is rare
> >
> > All the best,
>
>
> Thanks!
>
> Pitu.
>
>
> jens
> >
> >
> > _______________________________________________
> > olsr-dev mailing list
> > olsr-dev <at> olsr.org
> > https://www.olsr.org/mailman/listinfo/olsr-dev
> >
> >
> >
> >
> -------------- next part --------------
> An HTML attachment was scrubbed...
> URL: http://www.olsr.org/pipermail/olsr-dev/attachments/20060403/00bc0e52/attachment.htm
>
> ------------------------------
>
> Message: 7
> Date: Tue, 4 Apr 2006 08:53:36 +1000
> From: "Dan Flett" <conhoolio <at> hotmail.com>
> Subject: RE: [olsr-dev] auto ip address assignment
> To: "'OLSR development'" <olsr-dev <at> olsr.org>
> Message-ID: <BAY114-DAV8E785C461F30167585D18AFD50 <at> phx.gbl>
> Content-Type: text/plain; charset="us-ascii"
>
> I strongly believe that a decent, working implementation of auto address
> assignment for OLSR will cause a revolution in the popularity of mesh
> networking.
>
> Imagine a pure "black box" mesh node - you just power it up and instantly
> you have a self-configured mesh node. Without needing any general or
> specific knowledge about networking - the intelligence is all inside the
> black box. Lack of knowledge and unwillingness to learn are at present
> barriers to participation in a mesh network. If it was made super-easy then
> the only barrier would be willingness to participate.
>
> Revolutions in the uptake in technology have been caused not by the
> invention of the technology, but by making the technology the most
> user-friendly possible. I'd love to see a mesh network node that simply has
> an on/off switch - making it a household appliance and easily understood by
> the mainstream population.
>
> Cheers,
>
> Dan
>
> -------------
> View my blog:
> http://freenetjazz.blogspot.com <http://freenetjazz.blogspot.com/>
>
>
>
>
> _____
>
> From: olsr-dev-bounces <at> olsr.org [mailto:olsr-dev-bounces <at> olsr.org] On Behalf
> Of Pitu
> Sent: Tuesday, 4 April 2006 2:16 AM
> To: OLSR development
> Subject: Re: [olsr-dev] auto ip address assignment
>
>
>
> 2006/4/3, Jens Nachtigall <nachtigall <at> web.de>:
>
> > We've thought that using any existing implementation could be a good idea
> > (like PACMAN <http://sourceforge.net/projects/pacman-autoconf/>).
>
> I had a look at PACMAN a year ago, it is not working with the current olsrd
> release, also it only works on linux. But the ideas of pacman as described
> in
> the paper, are quite good, if I remember right. So if were you, i would read
>
> that paper and reimplement these ideas all from scratch in a olsrd plugin.
> The algos in the paper are enough to get it working, afaikr.
>
>
> Yes, it's the path that we would follow. PACMAN works as a daemon such that
> olsrd is not aware of his existence (as far as I'm concerned). The idea was
> to turn this code into an olsrd plugin in order to work closely with olsrd.
>
>
>
> Please also note, that nobody uses the original olsr-rfc-protocol, but
> rather
> the enhancements like etx. I think in the freifunk firmware, which most
> meshes use, also the MPRs are switched off, i.e. every node behaves linke an
> MPR. So the hierarchie is flat. This is because, the MPR idea did not work
> out.
>
>
> I didn't know that... it sounds quite frustrating. MPR seemed to be a good
> idea...
>
>
>
> Good luck on implementing. Rather get started, there are already so many
> papers on meshes, but code is rare
>
> All the best,
>
>
> Thanks!
>
> Pitu.
>
>
>
> jens
>
>
> _______________________________________________
> olsr-dev mailing list
> olsr-dev <at> olsr.org
> https://www.olsr.org/mailman/listinfo/olsr-dev
> <https://www.olsr.org/mailman/listinfo/olsr-dev>
>
>
>
>
>
>
> -------------- next part --------------
> An HTML attachment was scrubbed...
> URL: http://www.olsr.org/pipermail/olsr-dev/attachments/20060404/4605199c/attachment.html
>
> ------------------------------
>
> _______________________________________________
> olsr-dev mailing list
> olsr-dev <at> olsr.org
> https://www.olsr.org/mailman/listinfo/olsr-dev
>
>
> End of olsr-dev Digest, Vol 20, Issue 2
> ***************************************
>
RSS Feed