Cell Relay Archive

[craigp@world.std.com (Craig Partridge)]

"Brian Lavallee, P.Eng." <brian_lavallee@nt.com> writes:

>Shun Yan Cheung wrote:
>> 
>> Take a message of N bytes, and package them into cells
>> using AAL5 and other AALs. After doing this, count
>> the number of cells you used in each case.
>> 

>Is this supposed to help me?

There are two AALs for datagram encapsulation.  AAL3/4 places 4 bytes of
overhead in each cell, plus an additional minimum of 8 bytes of overall
overhead for each datagram.  AAL5 puts no overhead in each cell (it uses a
bit in the cell header) and adds an 8 byte trailer to each datagram (with
the rule this trailer must be at the end of the last cell).

So crudely speaking, the cost in cells of sending a datagram of length L
in AAL3/4 is:

    ceiling [ (L+8)/44 ]

and AAL5 it is:
    
    ceiling [ (L+8)/48 ]

Furthermore, AAL5 is more efficient to process than AAL3/4 (one need only look
at a bit in the header in each cell, rather than paw through the data contents)
and may well have better error detection properties.

Craig

PS: I thought Shun Yan's reply was indeed helpful.  One of the nicest short
summaries of the benefits I'd seen.