Cell Relay Archive

[jbasrur@sahana.bellcore.com (33150-basrur)]

> joytin,
> 
> aal 1 provides cbr service only, so there is no cell delay variation.
> as far as the buffer size goes, i suppose that you could look at the
> atm forum document on the "circuit emulation service" which uses aal
> type 1.  aal 1 is also based upon fixed aal-sdu size so you don't
> have a "buffer size" parameter as you would with an adaptation layer
> such as aal 3/4 which supports variable sized sdu's.  as i recall,
> the service specification for structured ces includes a "block size"
> parameter (where the block size is the size of an aal-sdu.  you would
> want a buffer of at least the size of the block.
> 
> as far as cell delay variation in general, the last time that i
> checked (which was earlier this year) there was no standardized
> method for handling cdv across atm networks.  as far as a
> relationship between tdev and cdv, i don't know anything about it.
> 
> 

I understand AAL1 provides CBR service only. But there is always a certain 
about of cell delay and delay variation in an actual network.  
especially when there are multiple
switches with varying port utilizations.
Regarding the buffer, the Convergence Sublayer does indeed have a buffer.
The SAR-SDU is placed in the CS receiver buffer. The CS does some processing
on these SDUs. It has to take care of the delay variation that has occurred.
This means that irrespective of how much variation has occurred it has to send 
the data at a fixed rate to the user.
 This however gives rise to some amount of jitter
and variance in delay. This cannot be characterized simply by the classical 
variance since the processes that give rise to phase variance are not
stationary processes. These processees are characterized by A and mu law 
functions. The TDEV parameter is used instead which converges for large samples. I am interested in what work has been
done in order to relate the TDEV and CDV parameters.
Thanks for your help.

Jyotin