Re: [STDS-802-3-EPOC] [802.3_EPOC] MMP implementation issues: Multiple FCUs and MMP, is it different?
It is possible to do grouping at CNU level(aggregate shaping for all
LLIDs of a CNU) at OLT since OLT knows CNU, but grouping of multiple
CNUs(belonging to a single FCU in this case) may have to be done
elsewhere as Ed is saying, however transmission is not guaranteed to be
in a single burst in this case as OLT bins the shaped group into
different queues(ex: CNUs).
This results in interleaving of bursts belonging to different profiles
w.r.t CoAX. It may be ok depending on how large the burst to each CNU
is.
But control frames(ex: MPCP grants/OAM) are locally generated and not
subject to the group shaper and transmitted independently, this could
result in short FEC termination that Ed was talking about yesterday.
-Venkat
From: Ed (Edward) Boyd [mailto:ed.boyd@xxxxxxxxxxxx]
Sent: Wednesday, January 09, 2013 11:08 AM
To: STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx
Subject: Re: [STDS-802-3-EPOC] [802.3_EPOC] MMP implementation issues:
Multiple FCUs and MMP, is it different?
Hi Andrea,
It is not guaranteed that the OLT can do the group shaping limit that
you describe below. Because of requirements in other system standards,
I think that all OLTs can shape to a single lower aggregate rate for all
downstream traffic. While a single limit is fairly common, the ability
to have 3 virtual pipes with aggregate limits as you describe below is
not as common but it is possible on some of the OLTs. It is likely to
be a shaping function in the switch or traffic management chip in the
box. Using existing solutions, an operator could specify that it has
group shaping for some number of pipes (3 in your example) and it seems
reasonable that an OLT vendor could provide the functionality based on
existing chips & systems. Requiring this function would reduce
selection set of OLTs to the operators but solutions do exist.
On the other hand, the MMP shaping function required is not a standard
function and it is far too complicated to implement behind the OLT MAC
with off the shelf chips. I don't see how it could come up the proper
FEC overhead for switching between modulation profiles and credit back
differently per destination for the fluctuating rate. You would need to
add this shaping function into the OLT MAC/EPoC PHY and then
back-pressure the switch in some way. This is a major change.
Hope that helps,
Ed..
From: Garavaglia, Andrea [mailto:andreag@xxxxxxxxxxxxxxxx]
Sent: Wednesday, January 09, 2013 7:04 AM
To: Ed (Edward) Boyd; STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx
Subject: RE: [802.3_EPOC] MMP implementation issues: Multiple FCUs and
MMP, is it different?
Hi Ed, all,
I still see one point that seems to me not fully clarified, talking
about SMP (simplest case) and a deployment like the one shown in the
figure.
Let me try to make an example, and see how this is supposed to work, for
my understanding:
- EPON is 10 Gb/s and serves 3 FCUs and 2 ONUs
- Each ONU is compliant to 802.3av and can handle 10 Gb/s speed
- The FCU has an optical side also compatible with 802.3av, but
cannot handle on the coax side 10 Gb/s - I think this is reasonable
assumption
o Let's assume the following for FCUs, on coax (other example won't
change the problem): FCU-1 runs at 2 Gb/s, FCU-2 runs at 1 Gb/s, FCU-3
runs at 1.5 Gb/s
o In each coax branch we use SMP as stated below, so CNUA, A', etc.
can handle a link from FCU -> CNU of 2 Gb/s, CNU B, B', etc. handles 1
Gb/s and CNU C, C', etc. handle 1.5 Gb/s
- In my opinion the OLT shall not:
o Exceed 10 Gb/s delivery to any ONU - this is already guaranteed as
being the upper limit of the EPON optical system
o Exceed 2 Gb/s aggregated traffic to FCU-1 - this is not guaranteed
by EPON line rate alone
o Exceed 1 Gb/s aggregated traffic to FCU-2 - this is not guaranteed
by EPON line rate alone
o Exceed 1.5 Gb/s aggregated traffic to FCU-3 - this is not guaranteed
by EPON line rate alone
It seems to me that the OLT needs to shape traffic for each FCU
according to the rules above, otherwise some of the FCU coax part will
be overloaded - so it seems to me the OLT needs to shape traffic to each
FCU and this is regardless the FCU particular architecture.
I assume the OLT shall and can do that - can be confirmed the OLT is
able to do that?
Thanks,
Andrea
From: Ed (Edward) Boyd [mailto:ed.boyd@xxxxxxxxxxxx]
Sent: Tuesday, January 08, 2013 19:59
To: STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_EPOC] MMP implementation issues: Multiple FCUs and
MMP, is it different?
Matt,
Thanks for the reply. It doesn't change anything. You can say that
A+B+C is more than 10Gbps but you won't be able to fully utilize the
coax networks. The key point is that the shaping rate for sub-network
(A, B, C) is a solid limit for each network. This is fairly easy to do
with a switch.
The shaping equation is very different if MMP is allowed on the network.
What is my bandwidth limit for network A if it has a mixture of modem
types? I set the shaping rate based on 1024QAM or 4096QAM? I assume
that you limit it to 4096QAM and decrement more credits if it is to a
1024QAM destination. Jorge's drawing suggest that we have a limit for
1024QAM modems across all networks A, B, and C. That doesn't prevent
congestion on any one network. The shaping needs to be smarter than
that. It would need to hierarchical (independent for A, B, and C) and
it would need to take into account the MMP data rate for each CNU on the
network. The more complicated decision is the FEC overhead from a short
termination. How does the RS layer know when the PHY is going to insert
a shortened last code word? I think that you would need to go back to
the packet FEC that was used in 802.3ah.
The rate control is very simple for the RS layer in the current plan.
We insert idles for a fixed FEC overhead for a fixed data rate pipe. It
is programmed value based on the PHY Link up result. When you talk
about MMP, I need to insert idles based on the spectrum configured,
per-LLID/MMP modulation order, and insert idles based on the shortened
FEC termination. The idles to insert for FEC is based on whether the
packet is in the same FEC block on the same coax network. This is very
complicated function and not available on any OLT. Jorge's drawing is
not accurate for that reason. There is not OLT that can perform this
function. Additionally, I think that it is very complicated or
impossible for the RS to come up with this function for the FEC. It
would need to be aware of the sub-networks and align its short FEC
terminations with the PHY.
Thanks,
Ed...
From: Matthew Schmitt [mailto:m.schmitt@xxxxxxxxxxxxx]
Sent: Tuesday, January 08, 2013 9:13 AM
To: Ed (Edward) Boyd; STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_EPOC] MMP implementation issues: Multiple FCUs and
MMP, is it different?
Ed,
Thanks much for putting this out there, as it helps illustrate where we
may be making different assumptions.
For me, I was not assuming that the sum of A+B+C is less than or equal
to 10 Gbps; on the contrary, I've been assuming that the total available
bandwidth for the sum of the FCUs would likely be greater than 10 Gbps.
If you change that assumption, does that change your conclusion about
the difference between multiple FCUs and MMP?
Thanks!
Matt
From: "Ed Boyd (Edward)" <ed.boyd@xxxxxxxxxxxx>
Reply-To: "Ed (Edward) Boyd" <ed.boyd@xxxxxxxxxxxx>
Date: Monday, January 7, 2013 5:32 PM
To: "STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx" <
STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx>
Subject: [802.3_EPOC] MMP implementation issues: Multiple FCUs and MMP,
is it different?
Jorge, Duane, and all,
I started a new thread since I wasn't quite sure where to insert my
reply. Let me start with the drawing with multiple FCUs question that
you proposed.
The assumption with multiple FCUs is that you are shaping 3 pipes (A, B,
& C) inside the 10G EPON downstream. If you make that sum of those 3
less than the maximum rate, you can send them down the PON at the same
with a single packet of jitter. The total bandwidth of the downstream
is a constant since it is the sum of the 3 constants.
For Multiple sub-pipes: Total BW = A + B + C (Sum of 3 Constants)
In the case of MMP, you have a single pipe with a changing data rate
based on the destination. The data rate from the OLT is not a constant
or the sum of 3 pipes. It is going up and down based on the
destination. A single shaping bucket is used but the debit for the
reverse leaky bucket is different based on destination. The total
bandwidth equation is an OR.
For MMP: Total BW = A | B | C. (The total bandwidth or maximum
bandwidth is based on the traffic)
I hope that helps. I will send a separate email on the other issues
related to the MAC layer solution.
Thanks,
Ed...
Edward Boyd | Sr Technical Director
Broadcom Corporation | (O) 707-792-9008 | (M) 707-478-1146
________________________________
<="" p="">
________________________________
<="" 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
