The IP over ATM Mailing List Archive by date

[onvural@vnet.ibm.com]

*** Resending note of 03/29/96 16:04                                           
 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -                 
 From:                                                                         
 Subject:                                                                      
Subject: Re: Mpt-Pt AAL5 delimiting                                            
Raif, Bert                                                                     
                                                                               
>Why mpt-pt (or mpt-mpt) ATM connections are required?                         
 You answered it well. Few other reasons have been posted already as well      
                                                                               
>VCI multiplexing. You can set up a mpt-mpt VP connection and assign a         
>unique VCI to each sender station. In that case the network does not          
>support VC connection, but only VP connections. When a station is             
>powered up/down no action is required but the notification of the VCI         
>to be used to that station only (note that this identifier might also         
>be locally derived, e.g from the less significant bit of the IP               
>address). A receiver can demultiplex concurrent incoming packets which        
>may be tranported using normal AAL5.                                          
 To me, this is the best solution that takes full advantage of ATM.            
 Multicast server approach could be equally good if the network is             
 (can be) provisioned accordingly. Current MCS work, I expect, will            
 solve various problems.                                                       
>The main disadvantage of this solution is that it is not applicable if        
>you use the ATM sub-layering for other purpose. For example, if you           
>create a private ATM network joining ATM LAN through a ATM VP pubblic         
>service (WAN) you cannot apply VCI multiplexing.                              
 Not really. Just use another VP if so desired.                                
>AAL multiplexing. You have to provide demultipling information inside         
>the cell. AAL 3/4 MID is an example, but in my understanding AAL 3/4          
>has not been considered suitable for our purposes because of its              
>complexity (e.g. CRC computation for each cell). Shall we need an             
 The main issue is the effective link utilization (44/53).                     
 Not necessarily the CRC computation.                                          
>AAL5+? [E.g: the first 32 word provides MID (maybe derived from the IP        
>address) and packet Begin/End indication].                                    
This will not solve the problem. You still need to wait for the whole          
packet to arrive. You either have MID in every cell or do a packet             
based approach. I guess there will be additional mechanisms that               
will be proposed in the near future.                                           
Regards, Raif