Cell Relay Archive

[ays@hooked.net (Alan Y. Schaevitz)]

In article <34F5E55A.8D02F338@txc.com>, Kaustabh Duorah <duorah@txc.com> wrote:

">"<P>I read through your earlier explanation and it seems to me that PPD
">"and EPD are handling slightly different conditions and both could be
implemented
">"simultaneously without a problem.
">"
">"<P>From what I gather. PPD will be used to discard the remaining part of
">"a packet when some cells are discarded somewhere up the line. This does
">"not have much to do with the congestion on the switch doing the PPD but
">"is used to increase the efficiency of the system. Off course it ultimately
">"helps to control congestion, I suppose.

Your description of PPD is not correct.  It would be impossible for a
switch to recognize that a prior switch had discarded a cell.  The only
indicator, the low order bit in the PTI field, is set to 0 to indicate an
cell that is part of a larger AAL SDU and to 1 to indicate the last cell
of that SDU.  Thus, if a single cell had been discarded, there would be no
indication of that to any subsequent switch.  PPD and EPD are variations
of the same procedure.  EPD discards an entire AAL SDU when it can detect
the entire SDU - that is, when it detects the "1" (i.e., last) cell of the
previous SDU on that virtual connection.  It can then discard all cells on
that virtual connection up to, and including, the next "1" cell.

However, if a switch must discard cells and does not know the "beginning"
of an SDU, it can still discard "0" cells up to, but NOT INCLUDING the "1"
cell.  Thus, some cells of that SDU will arrive at the destination where
they will be discarded when the SDU is reconstructed and fail the bit
error test.  This is called PPD.  The reason the last ("1") cell is not
discarded is because that would cause the receiving AAL to concatonate a
partial SDU with the next full SDU, causing the discard of two SDUs
instead of just one.

Regards,
Alan

-- 
Alan Y. Schaevitz
AYS Associates
ays@hooked.net