Shawn,
I’m probably not communicating clearly enough. You think in terms of statistics, operators typically plan deployments ahead of time. “5% of consumers can be served ONU’s with 25G capacity and 75% of consumers must have ONU’s with 50G and 100G capacity” is also a very odd way of looking at the problem – it is not number of customers, but amount of bandwidth sold to customer that matters. I can have 10 customers connected, each consuming 2 Gbps each, amounting to 20 Gbps of capacity, one customer needing 40 Gbps and one needing 55 Gbps and I would have to deploy accordingly. Thinking about a system like this only in terms of connected customers or billboard speeds is very limiting.
At this time, we can only speculate what future deployments will look like. I know what strategies are used today for 1G, 10G and migration between these two generations, and I have every reason to believe the same strategies would apply in the future. If that is the case, it will take groups approval to focus on 25Gx4 system as one of options for development. I have my opinion about it, as you correctly observed, and would rather we focus on what we originally set out to do, and not collect more requirements and system options.
Regards
Marek Hajduczenia, PhD, CCNA CSCO12874393
Network Architect, Principal Engineer
Bright House Networks
From: Shawn Esser [mailto:shawn.esser@xxxxxxxxxxx]
Sent: Monday, March 07, 2016 4:59 PM
To: STDS-802-3-NGEPON@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_NGEPON] 25G ONU Options to increase Network Capacity
Some comments on Marek’s input:
Re. 1) – the statements on flagship speeds apply really only to residential deployments. In terms of business services, there are no flagship speeds, and we deploy any data rates that make economic sense on the given platform. The evaluation of a breakpoint is typically done by comparing TCO for a P2P solution providing similar capacity, i.e., when it becomes cheaper to offer a specific number of Gbps to customer over dedicated fiber versus using shared medium. TCO in this case includes fiber cost, edge router cost, etc. to make sure that the comparison does make sense. From my perspective, the assumption in point 1) is false. The driver is to have ability to offer more than 25G in the future per customer, and that was the reason why we started looking at 50G and 100G options, meaning 50G and 100G MAC speeds, not aggregate for 4 independent channels. That is what was written into our objectives.
· No one is stating not to specify ONU’s with 50G and 100G MAC speeds.
· Not sure, but you may be stating that the hypothesis that most consumers can be served by ONU’s with 25G capacity is False. If so, this implies that you believe only 25% of consumers can be served ONU’s with 25G capacity and 75% of consumers must have ONU’s with 50G and 100G capacity.
· Jorge did not state that his calculations for flagship and billboard speeds were only limited to residential services but maybe that is what he meant. I do understand that business services need guaranteed service levels. Not sure if SLA for businesses would be the same as flagship or billboard data levels, or something less.
· In 10 years, what is Brighthouse’s expectations for % of NGEPON customer being residential versus business? If residential is a significant % of NGEPON deployments in 10 years, then these calculations still matter.
Re 2) – there is a lot of speculation as to where tunable optics will go. It is reasonably well established in long haul transport today, but even there cost delta is still there, after all the years of development and the reasons for deployment are different than in access. For some reason there is an underlying assumption that we need to build 4x25G system, that looks and feels like NG-PON2. I do not know where it came from and I do not believe this is what we were discussing in SG phase. I would be concerned that proving the need for such a project and its unique market identify would be rather hard, given that NG-PON2 is in development.
· Correct that Wavelength-tunable optics are well-established in Telecom long-haul. The premium for wavelength-tunable DWDM optics for long-haul is down to 20% or less over fixed-wavelength DWDM optics. This is for 88 wavelength channels in Telecom long-haul versus only 3-4 wavelength channels for NGEPON so theoretically a lower premium could be achieved for NGEPON. At high volumes, it is very feasible that wavelength-tunable optics for PON applications could also achieve 20% premium over fixed-wavelength optics. This means that instead of 3.5X premium over λ0 per Glen’s information, the premium could be 1.8X over λ0.
At this time, I would suggest we focus on delivering wat we set out to do: 25G system, 50G system, and 100G system – there are a lot of challenges without adding extra options to address undocumented requirements. Yes, we can bake in a lot of bells and whistles into this system, but the end goal is simple: low cost, scalable system, with typical Ethernet cost curve and reliability.
· These are not lots of bells and whistles. The options we are discussing are minor items to the standard: allowing 25G ONU’s to be on 4 wavelength pairs and providing provisions for wavelength-tunability. The former can be described in a couple sentences in the standard. Provisions for wavelength-tunability have been done before so my guess is that it could be borrowed from these other sources.
· No one has stated not to focus on delivering 25G system, 50G system, and 100G system. Adding these small provisions will not remove focus. We probably could have written the provisions already instead of all this correspondence.
· Allowing for these options can provide lower-cost and more-scalable systems.
Marek has made his views clear. We need to hear the viewpoints from the other cable operators. In the end, we need to take a vote and move on.
Thanks,
Shawn
From: Marek Hajduczenia [mailto:marek.hajduczenia@xxxxxxxxx]
Sent: Monday, March 07, 2016 3:33 PM
To: STDS-802-3-NGEPON@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_NGEPON] 25G ONU Options to increase Network Capacity
Nobody really does calculations for “average” speed on PON for business. The typically methodology is to add customers to PON as long as capacity allows, and at some deflection point look into either a) splitting PON into a smaller serving group, or b) migrating larger bandwidth consumers to higher speed PON. To put it into perspective, it is very common to have different PON branches with very different number of customers, ranging from 2-3 to 16-20 or more, all depending on the customer density in the given area. Residential builds are more predictable in that number of homes passed can be to some extent predicted and accounted for in capacity design. Business services do not have really this luxury.
I do not agree with the statement: “more cost effective to be able to add additional capacity beyond that of the first 25G by being able to allocate ONUs to additional instances of 25G” – this might work for future residential access, but it makes very little sense for business services, where capacity growth is much faster than in residential, primarily because of types of provided services, as well as SLA profiles. In fact, if we were locked down to 25G in the future, it would bring the need for a new standard sooner, which defeats I believe one of the requirements for scalable architecture up to 100G.
Regards
Marek Hajduczenia, PhD, CCNA CSCO12874393
Network Architect, Principal Engineer
Bright House Networks
From: Francois Menard [mailto:fmenard@xxxxxxxxxxx]
Sent: Monday, March 07, 2016 2:36 PM
To: STDS-802-3-NGEPON@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_NGEPON] 25G ONU Options to increase Network Capacity
Marek said:
Re. 1) – the statements on flagship speeds apply really only to residential deployments. In terms of business services, there are no flagship speeds, and we deploy any data rates that make economic sense on the given platform.
I agree that ‘Flagship & Billboard’ speeds are ‘residential’ concepts, wherein non-guaranteed & temporary peak speed have a perceived value.
I understand that business services are dimensioned according to an SLA based on ‘Sustained Average Throughput’ on the PON.
When 32 customers share a single 100 Gbps pipe, the average throughput per customer is 100 Gbps / 32 = 3.1 Gbps.
For some reason there is an underlying assumption that we need to build 4x25G system, that looks and feels like NG-PON2.
By being able to effect channel bonding at the PCS layer, the MAC rate can be greater than that of a single Ethernet lane. So it may be easier to argue that the 4x25G is more of a 4x25 Gbp PCS Lane system than a 4x25G MAC rate system. The idea is that when there is a need to burst at a faster rate than 25G, two PCS lanes of 25G can be bonded into a 50G MAC rate and four PCS lanes of 25G can be bonded into a 100G MAC rate. I understand that the are that in this case, should these lanes need to make use of λ0 for instance, than no 25G ONU at λ0 are allowed to burst when a 100G ONU is monopolizing all PCS lanes while transmitting.
It would seem to be much more cost effective to be able to add additional capacity beyond that of the first 25G by being able to allocate ONUs to additional instances of 25G, rather than make it mandatory to make use of 50G & 100G ONUs.
I am concerned with the limitations that would stem from NG-EPON having only a single channel of 25G and requiring that any expansion of capacity beyond 25G necessarily be done through the addition of 50G ONUs with a 2.5X relative cost or 100G ONUs with a 5.5X relative cost.
This would mean that any expansion beyond 25G would be occurring at greater cost than which could be afforded by the transition of Tunable ONUs from their current 3.5X relative cost to a plausible 1.5X relative cost between now and such time as 100G EPON becomes deployed by operators.
-=Francois=-