Re: [ns] Unicast and mcast in same simulation

[Tue Olesen <tue@cs.auc.dk>]

Hi!

We are two Ms.C. students currently working on modifying the existing NS2
implementation to support multicast-enabled wireless nodes. We have not
been involved with ns2 development before we started this piece of work a
couple of weeks ago. (For this reason, some of the arguments below might
contain errors due to lack of complete understanding of ns. Please correct
me if this is the case :-)

Comments to your posting are included below:

By the way: The purpose of this work is to enable simulations of multicast
protocols in wireless ad-hoc networks.



On Mon, 12 Nov 2001, Per Olesen wrote:

> 1) Are there any trouble in running both a unicast and a multicast protocol
> in parallel aside eachother in ns2? I can't think why there should be, but to
> be sure that there are not some hidden problems there, I ask anyway.

In wired network scenarios, it is not a problem to run multicast and
unicast routing side-by-side. Wireless nodes, though, does (according to
our knowledge) not have support for multicast traffic. When this issue has
been resolved, our guess is that this should not be any problem.

> I'm building my multicast extension on the assumption, that I can access
> tables maintained by the unicast protocol.

This assumption seems to be both realistic and pretty useful. We have
decided upon the same strategy ourselves.

[SNIP]

> 3) The unicast protocol implementation I'm building upon, is done all in C++.
> Doing a new of the protocol agent for a node, starts the implementation in
> C++. I fully understand that.
> In contrast, as far as I can see, all the multicast implementations contain a
> fair amount of OTcl code. Is this because we need to inherit McastProtocol,
> and that McastProtocol only exists in OTcl code?
>
Yes. The existing multicast protocols (DM, ST, ...) are all implemented as
subclasses of McastProtocol. The OTcl code performs, among other things,
updating of the multicast classifier and replicators responsible of
distributing multicast traffic. This task seems to be easiest to perform
in OTcl, but it should be possible to strip a new McastProtocol subclass
for your routing protocol down to contain a minimum of functionality, and
do the implementation of multicast routing logic and agent in C++.



Hope this helps!

Best regards,

Tue

-- 
Tue Olesen                   | mail: tue@cs.auc.dk
25, Rue des Bourdonnais      | phone: 0033 (0)6 98 09 88 82
78000 Versailles, France     | homepage: http://www.cs.auc.dk/~tue