[Commotion-dev] interesting tech offer from IRC

[L. Aaron Kaplan]

On Dec 15, 2011, at 10:24 PM, advice at csmctmto.bizfu.net wrote:

>> Was it tested in any larger network ? I.e. at least a couple of hundred nodes?
> 
> Tested, no.  Mathematically analysed, yes.  Simulated, yes.  It never got
> traction (a number of reasons, including nervousness of official
> intervention in what amounted to an illegal communications service).
> 
> The outcome: it provides an efficient routing solution out to at least a
> thousand routing nodes and an almost arbitrary number of leaf nodes, and
> given caching behaviour with any kind of stability to the topology, almost
> arbitrary efficiency after reorganization.
Sounds to good to be true. My experience is that there is always a trade-off between 
different aspects of a MANET protocol (for example: convergence speed vs. messaging overhead vs. available airtime)
> 


Ok, well my experience with OLSR-NG (www.olsr.org) is that you absolutely need to test code in a real world scenario! There are so many papers out there on mesh networks and MANETs and wireless networks that it became almost a whole branch of computer science ;-)

I recommend reading: "MANET studies - the incredibles" 
(http://citeseer.ist.psu.edu/viewdoc/summary?doi=10.1.1.110.7902)

Here is a list of MANET protocols - most of them theoretical:
http://en.wikipedia.org/wiki/List_of_ad_hoc_routing_protocols

The lesson is that most theoretical works need to reduce (abstract) and therefore really miss out on essential parts of Wi-Fi networks. Wi-Fi is actually rather complex. And radio waves makes it even more complex. Most simulations don't capture that.In addition many researchers are from the CS field and completely ignore layer 1. But good layer 1 in a collision zone is - well, little bit tricky.

In other words - when having a real world testbed, your nose is literally pushed into the messy dirt of reality. Every bug in the software makes you *walk* over to a node. Every bug costs you time and hence you really start to write very clean code (as well as think about theoretical aspects).

We (the EU wireless community folks) are now in the process of setting up a "planet lab" real world testbed for wireless mesh networks. It will be built in 2012 and will be open for researchers.
COMMOTION for sure could test things there.

Here is the web page: www.confine-project.eu


So much for my experience. And maybe it explains, why I am a bit skeptical.


Nevertheless , I'd be honestly happy to hear about any new ideas.
So, go for it, present your idea! :) We all collectively gain by the exchange of ideas :)


Aaron.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 163 bytes
Desc: Message signed with OpenPGP using GPGMail
URL: <http://lists.chambana.net/pipermail/commotion-dev/attachments/20111215/bf6ecb0a/attachment.sig>