Re: [STDS-802-3-EPOC] MMP ad-hoc - presentation for Jan.11 call
Hi Andrea & Nicola,
Thanks for the presentation. This presentation is very similar to the last presentation except for the polling diagrams that I'm not able to follow. Hopefully, I can make the next call so I can get a better understanding of the polling diagrams. In general, it argues the same points but doesn't contain numbers. I presented the best case and worst case penalty for shortening the code words. I think that others are now proposing a short code and a long code as an alternative. I would like to understand which you prefer. In either case, the profile switching penalty increases with the smaller pipe. I would like to see where our analysis differs. Additionally, if you aren't delaying the GATE frames, they will force shortened code words so that must be taken into account. It isn't enough to say that the penalty is small. Please add numbers so I can understand your assumptions versus mine.
We see differently on defining a scheduling function per LLID and backpressure to the 802.1 layer. Since it is not defined in the standard, you believe that you can define this scheduling function to support MMP. In my opinion, you are adding functionality above the PHY that doesn't exist in EPON MAC and is incompatible with the standard's compliant devices. I would also note that it isn't defined because it currently does nothing. Packet in and packet out. I think that it is a bad idea to add this scheduling function in the EPON MAC. The QoS and scheduling is handled by 802.1 before the EPON downstream. If we backpressure per LLID to the 802.1 layer, it is no longer performs scheduling. Eugene and I also pointed out that we can't come up with an idle insertion equation in the MAC for the PHY with MMP. It not clear how the MAC calculates the FEC overhead when codes are shortened. I would like to see your equation based on the design shown so I can better understand.
We still aren't connecting on the jitter issue as well. The shuffling of higher layer packets without regard to QoS in the MAC RS or PHY creates jitter for all services. It doesn't matter whether you reorder packets between one destination or two destinations. I understand that you aren't re-ordering to the same LLID. When you delay a packets differently from the input of the MAC on the CLT to the output of the MAC at the CNU, you have introduced jitter. The amount of jitter is based on the shuffling block size that has been introduced. I want to know what is your proposed block size for shuffling so the impact to MEF23H can be determined and the efficiency for the shortened code words can be calculated. Avi and I showed shuffling delays increasing to 4ms with the smallest pipe.
I'm also still not seeing where this comes together with your channel bonding proposal. Is it above or below it in the stack. I think that it is below but I'm not sure.
Thanks,
Ed....
From: Varanese, Nicola [mailto:nicolav@xxxxxxxxxxxxxxxx]
Sent: Friday, January 11, 2013 8:15 AM
To: STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx
Subject: [STDS-802-3-EPOC] MMP ad-hoc - presentation for Jan.11 call
Dear all,
Please find attached the presentation we prepared for today's MMP ad-hoc call.
Thanks,
Andrea and Nicola
________________________________
<="" p="">
________________________________________________________________________
To unsubscribe from the STDS-802-3-EPOC list, click the following link:
https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-EPOC&A=1