[Olsr-dev] OLSR and ns2 (or other sim)

Aaron Kaplan (spam-protected)
Tue May 6 16:33:21 CEST 2008


On May 6, 2008, at 4:18 PM, Chris Davies wrote:

> Aaron,
>
> If cross-layer optimizations are ignored (which ns2 wouldn't be really
> great at anyway), it would seem that proactive routing protocols such
> as OLSR need to do three things:
>
> 1. send/receive control packets
> 2. update routes
> 3. determine the current time (to figure out when to send packets)
>
> It shouldn't be too hard (having not done it of course :) ) to write
> an ns2 protocol that provides these hooks. Then olsr (or any other
> protocol) could connect using these hooks. It would require an extra
> abstraction layer on the protocol in order to be able to recompile
> olsr to use the ns2 hooks instead of the standard network stack/os.
>
> Any sense of if this would work? or am I missing something?

No, I think you are not missing any points. As far as I can see only  
of course that is :)

I was discussing this once already with somebody . But not for ns-2.  
But instead for his own simulator.
IMHO you would have to control the timers in olsrd and you would need  
some way to send and receive packets to some virtual network interface.
yes.

Be aware that currently (old codebase!) the "olsrswitch" program (not  
the windows olsr version) handles a maximum of 30 nodes until it  
crashes.
But it is a first example of how to do virtual networks.
For another example, also please take a look at the newer netsimcap  
code submitted by Henning. It also goes into that direction.
No matter what - for integration into any discreet time simulator you  
will need to send packets /recieve to a virtual network and control  
the timer (and have your own routing tables)

guys, sometimes I feel like a secretary who simply can tell you whom  
to contact or where you can find what ;-)
It has mostly been done, it just needs a lot of clean integration work.
That is the legacy of the olsr.org implementation - lots of old code.  
But layer by layer it is being rejuvenated ;-)))


>
> It would seem to be a good thing to be able to sim olsr easily during
> the development process in order to stress test it and more easily
> track down bugs.
>
absolutely!

care to sponsor some developer who would work just on that? (not me,  
to little time)
Also if time would not be a physical constraint, I would love to run  
coverity over olsrd.
Anybody has access to it? Done it? Willing to feed back the results ?

a.






More information about the Olsr-dev mailing list