Re: Modeling computation in nodes (again...I'm sorry)

[Yuri Pryadkin <yuri@who.isi.edu>]

Sounds OK to me.  The part about "busy flags" is a little bit unclear.
You didn't mention who is going to set/reset these flags.  Why cannot
you use a LinkDelay object instead (or a class derived from it)?  It
seems that by varying delay_ (and/or bandwidth_) you can achieve the
same thing.


  Yuri.

> I would like to extend the functionality of ns in respect to modeling
> the interleaving of computation and communication.
> 
> I read the documentation and am fairly certain that what I need to do is
> adding queues and busy flags to the Node (before classifier to
> model demux overhead), Agent (to model protocol overhead) and
> Application (to model application/service-specific computation).
> 
> Additionally I would need to extend the trace/monitor capabilities to
> observe these added queues and flags, I think.
> 
> Could you please give me feedback if that sounds like a feasible
> approach and/or if I forgot something which needs to be done too to
> extend ns in that way. Also, if you have any code or know some pointers
> to code that models computation in ns, I would be very grateful too.
> 
> Thanks in advance for your help.
> 
> Michael