[8023-10GEPON] Shortened FEC blocks
Dear Colleagues,
At the last meeting, we had a brief discussion on whether we need a
shortened last FEC block or not.
To illustrate the essence of the problem, consider a situation where an ONU
needs to send only one 64-byte message, such as a REPORT. If a FEC block can
be shortened, the ONU will send REPORT message and FEC parity data,
immediately following the REPORT. This approach also requires some kind of
FEC delimiters, to indicate where the protected data ends and parity starts.
Alternatively, if shortened FEC blocks are not allowed, the ONU in a similar
situation as above, would transmit a complete FEC block. If RS(239, 255) is
used, the ONU would add a padding to its 64-bytes of data to bring the total
number of payload bytes to 239, and then will add 16 bytes of parity.
However, this solution does nor require FEC delimiters, as parity data is
always found at a known location.
To quantify the possible overhead associated with not using a shortened last
block, I created an excel spreadsheet, similar to the one we used to
quantify PMD overhead in EFM.
Please, access the file here:
http://www.ieee802.org/3/av/working_docs/FEC_overhead_Ver1.xls
You may download a local copy and add additional columns for different
configurations.
As you can see from the charts, with 32 ONUs, in the worst case, not using a
shortened block introduces additional overhead of less than 0.4%. [The worst
case is considered when N-1 ONUs are idle (i.e, need to send only REPORTs)
and one ONU is very busy and would like to use all available bandwidth.]
This calculation ignored any additional overhead that may be caused by FEC
delimiters in case a shoretend block is used. Accounting for FEC delimiters,
will make the efficiency difference betweeen using a shortened block and not
using a shortend cblock even smaller.
Therefore, in my oppinion, it is not worth allowing a shortened FEC block.
Without shortened blocks system design becomes simpler and performance
impact is not significant.
I would like to hear more oppinion on this topic. Specifically, if someone
argues in favor of shortened blocks, please, suggest specific FEC frame
format, including FEC delimiters. With that information, I will be able to
modify the charts to compare effects of having delimiters to not having a
shortened block.
Regards,
Glen