So I decided to look at the scaling issue of PtP and came up with some very interesting results.
Fibre Termination
Being a Fibre to the Home Network the first and fundamental requirement of the headend/communication room is to terminate fibre. When looking at the various arguments put forward by PtP vendors, they never discuss how you terminate the 5000 fibres required to service the homes.
Despite the claims by some equipment vendors you can terminate 2000 fibres in a single rack, the maximum amount you can practically install into a rack and still maintain a level of maintenance and management is 576. This comes from the fibre contractors and ODF vendors themselves. Yes I can hear it now: "Corning has a big fibre management frame that holds 1440 fibres." Yes they do, but have you ever tried to get that many fibre cables into the frame, or even the patch cords. In my experience it becomes impossible to use after about half full.
For 5000 homes you need at least nine Optical Distribution Frames. This does not even consider RF Video overlay which requires a second fibre for PtP solutions because the equipment doesn’t support wavelength combining for the reticulation of RF TV for Foxtel.
Switching Equipment
The next point of consideration is the switching equipment where again you cannot effectively manage more than 500-600 fibres in a single rack; so the most optical fibre ports you can put in one rack is 600, again 9 racks required. While you can get two Cisco 6513s into a single rack and support 1056 subs, this number of patch cords into an ODF is a mess and impractical, let alone into a rack with equipment. Fibre needs to be carefully managed, and it is generally placed into a fibre ducting system which uses a “waterfall”. These are 50x50mm vertical ducts which extend down from the ducting system next to each rack. In each of these ducts the vendors recommend 306 is the maximum number of 2.0mm optical patch cords you can install and maintain some level of manageability. Using two of these ducts (one each side of the rack) means 600 fibres to a rack.
Power and Batteries
Next on the list is the power and batteries required to run this network; 5000 homes connected to nine Cisco 6513 switches would need 36kWatts of power. The heating for this is 113000BTU/hr requiring a 50kW air conditioner which equates to another 30kWatts of power. So all up you need about 66kWatts of power, which would require a direct 32kV power feed and a 100kW transformer (if not two). The batteries to run this and give a typical 8 hours of backup (what we design into all the sites we do for clients) equates to some 800 batteries, adding another 40 racks.
The monthly power bill alone would be over $7000.
Space
All up you need 60 racks to support the required equipment, fibre management and power systems. This would typically require about 150sq metres of floor space. This is no small transportable communications hut like the ones we recommend to our clients. Its more like a house which will have to be built in accordance with council guidelines requiring a block of 300sq meters. The communications building alone would be worth $200,000 (including the land), instead of the typical $25,000 for a 12sq metre transportable. Alternatively leasing a building would add another $5000 per month to the overheads.
Active Field Equipment
One alternative approach is to use a distributed model, whereby the switching equipment is installed into Fibre Access Nodes placed strategically throughout an estate. While this may address the fibre management issue, it still consumes the same amount of power, requires more air conditioning and has a considerably higher installation cost. A street node with no active equipment, costs about $30,000 to supply and install (without equipment); with nine nodes this would cost $60,000 more than a central communications room. It is also much more complicated to maintain and offers less backup time due to the limited battery storage space.
Conclusion
I have gone through this issue many times with PtP vendors in residential applications. PtP does not scale when you consider what I have listed above. It is physically doable but at what cost?
So to those Ethernet switch vendors out there who believe they have a Fibre to the Home soluton because they have a switch with 100baseBX ports; thing about what you are recommending to your clients and consider some of these points above.
6 comments:
- At 19 August, 2008 13:45 Ger Vloothuis said...
-
Did you talk with the people of Citynet in Amsterdam? They are doing 450,000 subscribers in P2P with 2 cores per subscriber.
Yes, it ain't easy. But they are doing it. - At 19 August, 2008 15:38 Stephen Davies said...
-
They are not doing 450,000 YET. They are planning to do 450,000 once the project has been completed, but there is a long way to go.
Europe is also very different to Australia, because of apartment style living which enables the installation of mini pops in the basement of each building. this results in the average loop length in Europe being only about 800m compared to the average in Australia of 3.5kms. - At 20 August, 2008 01:46 dwillson said...
-
The trade of between PtP and PON currently has been the cost of the CPE. Fiber connections have to be managed at the concentration point, whether this is a central location or at smaller distributed points.
If you concentrate the fiber centrally, yes you will need a fair amount of space and power. This is managable and is no different from today's Central Offices with masses of wire (or fiber) that has to be managed!!
If competitive service providers are to be encouraged and network access is to be "open", then in addition to managing the quantity of fiber (or copper) there has to be additional functionality in the system to "patch" the subscriber to their choice of service provider. If you "patch" remotely you want to do this automatically in a manner that does not require a truck roll.
Users (and regulators) should voice their requirements for "openess". This requirement translates to user being in control of access path rather the service provider. Today you are told by the service provider that you are connected to, what service you will get, when and at what price. You have no other choice!! Vendors design products for "near" monopoly service providers whose objective (quite naturally) is to maintain the status quo! Giving the user "control" will not reduce the amount of fiber but will require that the fiber (or copper) drop from the customer premise to the concentration/selection point is controlled by the user. - At 20 August, 2008 07:43 Stephen Davies said...
-
Open access FTTH does not need to occur at the physical layer as indicated here. It can more easiler and cost effectively exist at layer 2.
If as you suggest using open access at the physical layer, the patching of the fibre from one service provider to another will ALWAYS require a truck roll as the CPE will require replacement.
Using regulatory power to enforce incumbants to "open" their network and provide layer 2 wholesale access eliminates any need to visit the subscriber.
FTTH is not just about delivering fast broadband like xDSL networks. Its about delivering a multitude of services including at least telephony, internet and PayTV.
Business models where you migrate the customer at the physical layer and send out a replacement "modem" are based purely on a Internet Service Provider model and not one which enable the increasing of ARPU by the introduction of newer or more advanced services.
In so far as the centralised splitting issue, while it MIGHT be fine for incumbants, few competitive carriers would have the space available. The cost of such is also prohibative with the space and power load requirements.
I might add a number of incumbant carriers are planning to reduce the number of exchanges they operate. For PtP this would be impractical.
As for the cost of the CPE, I have heard this arguement from the PtP equipment vendors and some pro PtP consultants for years and have proven this a total fallacy.
The CPE is only one part of the total network, and all costs must be considered when making a comparison, including the outside plant.
PtP solutions cost considerably more when both the active and passive equipment is considered.
I might add the old "PtP CPE is cheaper" arguement is now dead and gone. You can purchase indoor PON CPE at prices the same as indoor PtP CPE.
There is a white paper on this blog which makes the cost comparision of the two solutions taking into account Total Cost of Network ownership (including ongoing maintenance) - At 08 September, 2008 14:49 Anonymous said...
-
There is an elegant answer that this discussion missed. In an estate involving industrial, commercial and residential the splitter splices were located in the equipment room. This realises the termination savings Stephen mentioned, while each customer still has a dedicated fibre. Should any customers need 1Gbps connections the splice is opened and their cores connected via spares to a real termination.
The problem for PtP fibre is not the cores, but it is as Stephen says, the terminations. Dedicated cores costs about 1c per metre, which is almost negligible. This solution brings the best of both worlds. - At 10 September, 2008 21:52 Stephen Davies said...
-
Thanks for your input.
Your suggestion is quite valid and does provide a strong option enabling a very flexible network. We have used this approach in industrial estates by providing point to point fibre back to a mini exchange. The network operator can then use PON or PtP equipment.
An a typical industrial estate is likely to have less than 1000 premises, so two fibre racks would be more than ample to cope.
However the orginal artical was about a Fibre to the Home estate (not commercial or industrial) where the densities are much higher and the volume of subscribers and therefore fibre terminations much bigger.
It would be quite reasonable to combine PON throughout the residental section and PtP in the commercial section.
Alternatively using semi centralised fibre cabinets (services about 384 fibres) could also be used.
