Re: Ns forwarding support for multicast shared tree arquitectures

[Ahmed A-G Helmy <ahelmy@catarina.usc.edu>]

> Hi,
> 
> Ns support for multicast forwarding (multiclassifier + replicators) is
> source based oriented: a replicator is created for each pair <src,
> group> , resulting in an innecessary simulation overhead when simulating
> shared tree protocols such as CBT (in such case  is necessary create
> only one replicator for multicast group -and for each input interface).
> Is there any work to implement Ns support for both, source based and
> shared tree architectures?

There is support for an early version of detailed PIM-SM [i.e. the version 
does not implement complete PIM, but supports all shared tree mechanisms]
the support is given for both shared trees and source based trees... 
[tcl/pim/*.tcl]

the current implementation paradigm resembles that of mrouted and pimd in 
a unix implementation... i.e. the protocol holds (S,G), (*,G) ... etc, 
and the forwarding cache (at the unix kernel) [implemented in C++ in ns] 
holds S,G entries only.

You can definitely implement your own *,G classifier, based only on group 
address, however, the lookup mechanisms would probably be different
[for cbt , this is even easier than in PIM-SM, since cbt does not support 
source based trees, ttbmk].
at this point, I am not aware of any code supporting *,G forwarding entries,

Did you actually run into a memory problem with S,G forwarding states ?!
was that the bottleneck in your simulations ?!
I am interested to hear about your experience,

Regards,
-A

> 
> Regards,
> 
> Guillermo Rigotti
> Computer Science Departament
> Universidad Nacional del Centro - Argentina
> 
>