Re: Ns forwarding support for multicast shared tree arquitectures

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

> I am implementing CBTv2, except Hello Protocol (not yet implemented) and
> bootstrap mechanisms (not planned). Currently I am debugging CBT PDU
> interchange, without generating multicast traffic. The code includes periodic
> interchange of echo request and echo reply messages, timers specified by CBT
> (all implemented in OTcl) and debugging facilities (messages sent from CBT
> entities to standard output when an event occurs - PDU sent or received,
> timeout, etc).
> I use a small topology for debugging (4 subnets and 15 nodes -total-), because
> of  it I have not experienced any bottleneck  My question is related to the
> following subject:  Ns forwarding design is previous to shared tree 
> paradigms or
> Ns designers considered proper to implement (S,G) replicators as a general
> approach?.

S,Gs are one way where shared trees may be implemented, .. this is how 
a version of PIM-SM is implemented in ns currently.
The design was done with both shared and source-based trees in mind,... 
however, S,Gs *may* not scale very well if you have so many sources and 
groups (since it grows in O(S x G) )... 
but these entries are in C++ and likely to be efficient in space and time 
consumption (I have no quantitative data handy to substantiate this 
statement, though), in terms of per packet processing.

if you don't plan to run very large simulations, then I think the current 
paradigm may suffice... 
if S,Gs become the bottleneck then you probably can do *,G entries by 
inheriting from the S,G entries and indexing only on the group address.

hope this answers your question,

Regards,
-A

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