Cell Relay Archive

[Vinod Jeyachandran <vjeyachandran@mil3.com>]

Here are my observations on the behavior of AAL2.

  If the CPS of AAL2 follows the I.363.2 recommendations there will no 
requirement for a SAR layer. The CPS-PDU that comes out of the AAL2 CPS 
layer is exactly 48 bytes - no less no more.


(1) The Payload from the SSCS layer is received as the CPS-SDU. Assume that 
the Payload is greater than 44 octets
  ____________________
|                                   |
|____________________|

|---- CPS SDU --------------|

(2) A three byte CPS header is added to the incoming SDU and a CPS packet 
is formed. (Refer Sec 9.1). The TIMER_CU is started.
  ____________________  ___
|                                   ||       |
|____________________||____|

|----------CPS Packet ---------------|

(3) A 47 Octet segment is removed from this packet and a 48 Octet CPS-PDU 
is created by adding a One Octet header to this CPS-PDU Payload. (Ref (Sec 9.2)
  _____________  ___
|                       ||      |
|_____________||___ |
                                                              ===>   ATM 
Cell Payload (Pre-Adapted)
|--- CPS PDU ---|
       Payload
|-- CPS PDU ------------|

(4) The remaining octets of the CPS packet remain in the CPS layer until 
expiry of TIMER_CU or arrival of more data from the higher layer sufficient 
to create a CPS-PDU payload. At this point the remaining Octets are removed 
from the buffer and 48 octet CPS-PDU is created. Padding is done if required.

  The operations of the CPS-transmitter is fully described in section 10.1 
of the recommendation.

I believe that the rationale behind incorporating the SAR functionality 
into the CPS layer is to increase the channel utilization and at the same 
time keep the ETE delays low. Though the argument does not hold very good 
for large packets, it makes good sense when dealing with shorter sized packets.

Back to you guys.

Vinod

************************************************
Vinod Jeyachandran,
Senior Engineer.
MIL 3, Inc,
Old protocols never die, they just get in the way.
************************************************


>Here's my take on this extremely interesting posting (by Jeff Wilson)
>and
>the subsequent response (by Baris Aksoy).  KINDLY FEEL FREE TO CORRECT
>ME,
>SHOULD ANY OF MY OBSERVATIONS BE WRONG.  To keep this posting short and
>readable, I am splitting it into multiple postings.  Here's part II:
>
>
>Point 4
>=======
>I have often wondered myself whether there was a need to make the
>structure of AAL2 so radically different from other AALs.  Consider
>this (based on I.362 and I.363):
>
>AAL 1
>|---Convergence Sublayer (CS)
>|---Segmentation and Reassembly Sublayer (SAR)
>
>
>AAL 3/4, 5
>|---Convergence Sublayer (CS)
>     |---Service Specific Convergence Sublayer (SSCS)
>     |---Common Part Convergence Sublayer (CPCS)
>|---Segmentation and Reassembly Sublayer (SAR)
>
>If you make a tree structure from this hierarchy, and do a post-order
>traversal of that tree, you get a good indication of the various
>adaptation processing steps, in the correct order
>
>
>                                AAL
>                               /   \
>                              /     \
>                             CS     SAR
>                            /  \
>                           /    \
>                         SSCS   CPCS
>
>
>You begin with the service specific convergence operation [SSCS] (which
>is optional), following it up with a common convergence operation [CPCS]
>(where the service type is not discernable).  This completes the
>convergence operation [CS].  This is followed by the
>segmentation/reassembly operation [SAR], and once that is done, the
>entire adaptation process is completed [AAL]
>
>
>In AAL2, the sub-layer hierarchy makes a complete departure from I.362
>section 2.1, which states that:
>"To support services above the AAL, some interdependent functions must
>be performed in the AAL. These functions are organized in two logical
>sublayers, the convergence sublayer (CS) and the segmentation and
>reassembly sublayer (SAR). These sublayers may be further subdivided."
>
>
>For AAL2, the sublayer structure looks this:
>
>
>                                     AAL
>                                    /   \
>                                   /     \
>                                  /       \
>                                 /         \
>                                /           \
>                               /             \
>                              /               \
>                           SSCS               CPS
>                           /  \              /  \
>                          /    \            /    \
>                         /      \          /      \
>                        SAR   Trunking   CPS      CPS
>                        /|\              Packet   PDU
>                       / | \
>                      /  |  \
>                     /   |   \
>                    /    |    \
>                 SSSAR SSTED SSADT
>
>
>As you can see, my "postorder traversal" theory collapses for AAL2
>(or doesn't it?)
>
>To date, no one has been able to explain to me satisfactorily as to why
>such a radical departure from I.362, section 2.1 was necessary.  At the
>very least, a corrigendum should be issued for I.362, explicitly
>mentioning that this section does not hold true for AAL2.
>
>Flames welcome.....
>
>- ----Richard