[B.A.T.M.A.N.] batman - how does "fast internet connection" connection work

[B.A.T.M.A.N.] batman - how does "fast internet connection" connection work

[david johnson]

Hi

First an introduction - I am a David Johnson - I've been working on mesh
protocols for a few years now and I've built a 49 node wireless grid to
benchmark mesh protocols in South Africa:
http://wirelessafrica.meraka.org.za/wiki/index.php/49-node_Indoor_Mesh

Elektra and I did some experiments with batman (do I have to use the
dots @#!!!) and olsr at the beginning of the year and we wrote a paper
showing how batman outperforms olsr on CPU usage and better throughput
due to less flapping. The results were very impressive!

I'm now in the USA starting a PhD in mesh and I've decided to work on
batman in one of my networking projects. We have this idea to try and
improve gateway location performance by measuring real throughput at the
gateway and then allowing nodes to optimally pick gateways that will
give them the best throughput to the Internet but in a fair manner - ie.
one user musn't disadvantage another.

I would like to understand how the "fast internet connection" gateway
location node works
batmand -r 1 [interface]?:

It says it considers link quality and advertised gateway class - what
formula does it used to combine these two together so that it compare
others against each other.

Thanks and look forward to becoming part of the discussion

David

Re: [B.A.T.M.A.N.] batman - how does "fast internet connection" connection work

[aaron]

david johnson wrote:
> Hi
>
> First an introduction - I am a David Johnson - I've been working on mesh
> protocols for a few years now and I've built a 49 node wireless grid to
> benchmark mesh protocols in South Africa:
> http://wirelessafrica.meraka.org.za/wiki/index.php/49-node_Indoor_Mesh
>
> Elektra and I did some experiments with batman (do I have to use the
> dots @#!!!) and olsr at the beginning of the year and we wrote a paper
> showing how batman outperforms olsr on CPU usage and better throughput
> due to less flapping. The results were very impressive!
>   

Well, it is interesting that you note that.
I read the PDF. One thing that struck me was that the settings for OLSR
were such that OLSR used *full flooding*. (MPR 7 etc)
That is *exactly* NOT what it is meant to do!
So I must object to the objectivity of the study.
Basically the study was comparing an elephant and a wale
Well.. both are great ideas :) But ... they do live in different terrain
and under different assumptions :)))

got me?


> I'm now in the USA starting a PhD in mesh and I've decided to work on
> batman in one of my networking projects. We have this idea to try and
>   
great!
> improve gateway location performance by measuring real throughput at the
> gateway and then allowing nodes to optimally pick gateways that will
> give them the best throughput to the Internet but in a fair manner - ie.
> one user musn't disadvantage another.
>
> I would like to understand how the "fast internet connection" gateway
> location node works
> batmand -r 1 [interface]?:
>
> It says it considers link quality and advertised gateway class - what
> formula does it used to combine these two together so that it compare
> others against each other.
>
> Thanks and look forward to becoming part of the discussion
>
> David
>
> _______________________________________________
> B.A.T.M.A.N mailing list
> B.A.T.M.A.N@open-mesh.net
> https://list.open-mesh.net/mm/listinfo/b.a.t.m.a.n
>   

Re: [B.A.T.M.A.N.] batman - how does "fast internet connection" connection work

[Simon Wunderlich]

[-- Attachment #1: Type: text/plain, Size: 1774 bytes --]

Hello David,

On Wed, Oct 29, 2008 at 11:49:30PM -0700, david johnson wrote:
> Hi
> 
> First an introduction - I am a David Johnson - I've been working on mesh
> protocols for a few years now and I've built a 49 node wireless grid to
> benchmark mesh protocols in South Africa:
> http://wirelessafrica.meraka.org.za/wiki/index.php/49-node_Indoor_Mesh
> 
> Elektra and I did some experiments with batman (do I have to use the
> dots @#!!!) and olsr at the beginning of the year and we wrote a paper
> showing how batman outperforms olsr on CPU usage and better throughput
> due to less flapping. The results were very impressive!

Nice. :)

> 
> I'm now in the USA starting a PhD in mesh and I've decided to work on
> batman in one of my networking projects. We have this idea to try and
> improve gateway location performance by measuring real throughput at the
> gateway and then allowing nodes to optimally pick gateways that will
> give them the best throughput to the Internet but in a fair manner - ie.
> one user musn't disadvantage another.

That sounds interesting! Please keep us informed. ;)

> 
> I would like to understand how the "fast internet connection" gateway
> location node works
> batmand -r 1 [interface]?:
> 
> It says it considers link quality and advertised gateway class - what
> formula does it used to combine these two together so that it compare
> others against each other.

You can look it up at batman.c:393 (in the current trunk)
The formula used is:

  max( TQ^2 * download_speed ) 

(with a constant factor which is not relevant for the maximum, so i
skipped it)
TQ is the average transmit quality to the gateway, download_speed 
is its advertised download speed. 

best regards,
	Simon

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: [B.A.T.M.A.N.] batman - how does "fast internet connection" connection work

[Benjamin Henrion]

On Thu, Oct 30, 2008 at 7:49 AM, david johnson
<david.lloyd.johnson@gmail.com> wrote:
> Hi
>
> First an introduction - I am a David Johnson - I've been working on mesh
> protocols for a few years now and I've built a 49 node wireless grid to
> benchmark mesh protocols in South Africa:
> http://wirelessafrica.meraka.org.za/wiki/index.php/49-node_Indoor_Mesh

I would not trust that much indoor applications with so much reflections.

A tested like this might be interesting outdoor with large distances
between nodes (let say 500m between each node with 5dbi antennas).

-- 
Benjamin Henrion <bhenrion at ffii.org>
FFII Brussels - +32-484-566109 - +32-2-4148403

Re: [B.A.T.M.A.N.] batman - how does "fast internet connection" connection work

[Alexander Morlang]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

aaron schrieb:
> david johnson wrote:
>> Hi
>>
>> First an introduction - I am a David Johnson - I've been working on mesh
>> protocols for a few years now and I've built a 49 node wireless grid to
>> benchmark mesh protocols in South Africa:
>> http://wirelessafrica.meraka.org.za/wiki/index.php/49-node_Indoor_Mesh
>>
>> Elektra and I did some experiments with batman (do I have to use the
>> dots @#!!!) and olsr at the beginning of the year and we wrote a paper
>> showing how batman outperforms olsr on CPU usage and better throughput
>> due to less flapping. The results were very impressive!
>>   
> 
> Well, it is interesting that you note that.
> I read the PDF. One thing that struck me was that the settings for OLSR
> were such that OLSR used *full flooding*. (MPR 7 etc)
> That is *exactly* NOT what it is meant to do!

yes, the MPR idea was a good idea, in theorie.

as i heard, it did not work out at all and made the mesh even more unstable.

i think, the problem was in the way of selecting MPR, but i don't
remember exactly, elektra did a very good explanation that time, some 2
or 3 years ago.

it is quite sad to see that knowledge lost in the current OLSR developer
community.

elektra, could you be so nice and enlight aaron on the topic of problems
with the MPR selection?

> So I must object to the objectivity of the study.
> Basically the study was comparing an elephant and a wale
> Well.. both are great ideas :) But ... they do live in different terrain
> and under different assumptions :)))
> 
> got me?

please calm your temper.

...

>> Thanks and look forward to becoming part of the discussion
>>
>> David

Alex

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.8 (Darwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEARECAAYFAkkYO1QACgkQhx2RbV7T5aEaRwCfVw47gRH1mVfwVj0Qqxbh37e2
m7kAn1nT6v6EeMdQrJucI+iEs+grbZEl
=a2lu
-----END PGP SIGNATURE-----