Cell Relay Archive

[Ahmed Mehaoua <amehaoua@crim.ca>]

Toma Belichki wrote:
> 
> Why the ATM cell is fixed length and just 53 bytes long?
> 
> I know  the fixed length makes possible the fast hardware
> switching, but what about the length. If the cells were, lets
>  say 100 bytes, there should be much smaller overhead.
> I've hurt the length comes from the demand of  real time
> applications  such as voice and video, but how.
> Does somebody knows? ...
> 
>
	
	Dear Toma,

	This question have been discussed at numerous times.
	But as you know, ATM has been intially designed and promoted
	by Telecom operators (CNET France telecom labs, Bellcore) to convey 
	voice data.

	During the standardization process, the 48-bytes ATM payload type has
been
	a compromise solution between the european (32 bytes) and the american
(64
	octets) propositions.

	Voice transmission is an 64 Kbps isochronous application with
	two stringent temporal requirements :

	1- one byte should be transmitted every 125 micro-sec and
	has to be delivered to the destination codec at the same
	frequency rate. 

	2- the propagation delay should be bounded (<= 28 ms) to ovoid
	problems (echo, adaptation, ...) in long distance connection.

	Considering the second temporal condition :
	The total transmission delay (D) should be <= 28 ms.
	D is composed of three components :
	D= packetization delay + propagation delay + depacketization Delay

	Since the source should generate one byte every 125 micro-sec, 
	the packetization delay to fill a 48-byte cell is exactely 6 ms
	(48*125micro-sec = 6ms).
 
	The same time is needed by the destination to depacketize the
	cell (another 6 ms).  

	The remaining time for the propagation delay is 16 ms (28 - 6 - 6)
	to respect the overall bounded delay.	

	If you assume that the medium transmission rate is about 200,000 km/h,
	you can then deduce the maximum transmission distance : 3,200 kms

	For european continent, it is quiet enough. But in north america, echo
	concealment equipments have been widely installed to overcome this
distance
	limit.

	Since they are not limited by distances and related signal problems, 
	american operators have proposed to use a larger payload type (64 bytes
        or larger) to optimize the bandwidth utilization.

	In the other hand, european operators have proposed smaller sizes 
	(32, 16 bytes) to reduce the packetization/depacketization delays and 
	ovoid the need to install echo concealment equipments in their
networks.

	Consequently, for technical and economical reasons a 48-bytes payload 
	cell have been choosen.  	


	Hope this will help,
	Ahmed. M.

	
-- 
________________________________________________________________

Ahmed Mehaoua
Telecommunications & Distributed Systems
Computer Research Institute of Montreal (CRIM)

1801 McGill College ave, suite 800
Montreal (Quebec) H3A 2N4 CANADA
mailto:Ahmed.Mehaoua@crim.ca
________________________________________________________________