The IP over ATM Mailing List Archive by date

[Fahad A Hoymany <fahad@cs.pitt.edu>]

Raif Onvural writes:

> Subject: Re: Multicasting with ATM - a concern                               
> Another option is to use cell training. That is, all the cells of            
> a packet may be transmitted consecutively on the link.                       
> (imitating a packet with some extra overhead, but that is ATM).              
> All cells of a packet are collected at an intermediate node and             
> the transmission at the outgoing link starts after all the                   
> cells of the packet are received. Of course, minor optimizations             
> are possible.                                                                  

  This is the idea used by reassembly-mode connectionless servers (I.364),
but I think in "true" multicasting this does not work, since there is
no intermediate node.  To achieve what you suggest above means either
the switch will have to buffer the multicast packets and send them
in a serial order (fat chance) or you are basically going to impose
a multicast overlay network on top of ATM (as in CLS service) which is
not efficient.

Fahad
                                 %
Fahad Hoymany                   %%%
fahad@cs.pitt.edu              %%%%%
University of Pittsburgh      %%%%%%%
(412) 488 0404                   |