[ns] Re: Basic NS questions
Mon Apr 29 18:15:01 2002
IMHO, you need to implement your own MAC...should be able to use the
current node in ns...just a thought.
On Mon, 29 Apr 2002, Amar Chadha wrote:
> Hi Xuan,
> I hope you could spare a few minutes to answer some
> basic questions on NS. Just looking for pointers on
> how to get started. I've looked at the tutorials and
> the manual, however its not obvious how to approach
> the task -
> My task is to scope out how to model
> an RPR (resilient packet ring) node and then link
> multiple RPR nodes on a Sonet/SDH ring to simulate
> traffic policing and the congestion control
> fairness algorithm.
> Assumption - topology = ring (all point-to-point
> connections full-duplex). Physical medium.
> Basic questions -
> How does one go about trying to specify a new MAC
> device? The main aim is to first model an RPR MAC device.
> * Will it be necessary to define a new Mac class to
> incorporate the RPR MAC functionality similar to the
> way say the Mac/CSMA/Cd is defined.
> * Does it make sense for the MAC functionality to be
> incoporated into a new type of node - say RPRNode
> * The node will contain mutliple transmit queues
> with varying priorities.
> * In the node - A new packet format will need to be
> defined, so is it fair to say that a
> new address classifer and new recv() function will also
> have to be defined?
> Or do you suggest that the MAC functionality be implemented
> below the Node somehow.
> I'm finding it very confusing as to how to get started.
> Thanks in advance,
> Amar Chadha
> Multilink Technology
> Some details -
> The RPR MAC will sit in-between the Sonet framer (PHY)
> and the client MAC (most likely Ethernet MAC(s)).
> The RPR MAC will have the Transit/Add/Drop buffers.
> Will have to implement
> - congestion detection (based on the buffer (queue) threshold levels
> - congestion indication (generate fairness control frames)
> One of the tasks will be to determine the optimal buffer
> size and the queuing scheme for the Transit/Transmit traffic.
> Additionally - model the two fairness approaches and determine
> the settling behavior.
> Any pointers on what kind of new structures need to be created?
> Are the Queue objects sufficient
> Is there sufficient information or built/in statistical information
> that can be easily extracted from the queue models.
> I'm trying to figure out the basics without having to spend an
> enormous amount of time wading through the details.
> It seems to me that at a minimum we need to define a new -
> and possibly a new Queue.
> and all of the methods required for the new objects.
> Is this a fair assumption. Any idea how long it would take?
> Any help is appreciated.