Thursday, 11 December 2008

PON networks vs P2P networks

The discussion between Point to Point networks and passive networks with splits close to the end user gets very heated because of different vendors getting involved in the discussion and touting their wares. However if you look at it from the point of view of the owner of the network and the options they have for the future with the network, then the discussion becomes alot clearer and also easier.

If you're an incumbent like Verizon, with a near monopoly and you don't have to worry about regulatory intervention, access by third parties, competition etc, then and only then splits close to the end-user might not be too  much of a problem. You might save about 10 or 20% on the Capex, by keeping the network "fibre lean".

However if your position as the owner of the network is more like the incumbents and new entrants in Europe, then the world changes pretty quickly. The physical layout of the network shouldn't excluede any optical technology you want/can use. So you can use either GPON, EPON on a point to point network as well as plain standard p2p Ethernet (or even ATM, SDH, xWDM if you would be so inclined). With splits low in the network, you do exlcude any networking technology that can't handle a network where different users share the same line. A flexible, fibre rich network that allows the world to change over 50 years is the answer.

As always, where you sit is where you stand.
- Regulators should want P2P (shame on CMT for not requiring Telefonica to use P2P)
- Network infra players should want P2P
- Any investor with a 10+ year investment horizon should want P2P
- Mobile operators should want P2P
- Incumbents with fear of regulation, long investment horizons, visions of remainig relevant for another 50 years, multi-service incumbents etc should want P2P
- Infra players should want P2P
- Monopolist with nothing to fear and shortsightedness should want low split fibre networks.
- Vendor touting PON wares --> Low split may be nice as it 'guarantees' sales in the future. Or it might not, as in 10 years your selling different wares.
- Vendor touting P2P wares  ---> Easy

The choice then between splits low in the network or a point to point network, then becomes a choice between flexibility over the life-time of the network, the ability to quickly raise penetration rate on the network and strategic possibilities to include or exclude service providers. Reggefiber in the Netherlands probably said it best "Fibre is a real estate play". And something they understand but don't shout too much, fibre is also a first mover market. So looking at the three items that make a busines case:

1. P2P offers the most flexibility.
 a. It supports different use cases, different customers etc best. Any network will have to support any possible mix of customers. It's not only fibre to the consumer, its also fibre to the business and fibre to the wireless antenna site. (and who knows, fibre to the lamppost might become an option). P2P designs allow FTTB/FTTH etc. to be mixed on the the same physical network. Low split may not be acceptible for FTTB type applications as it is a shared network, making everyone dependent upon the same end-point and the same technology used a the end-point. Simply put, some companies already need 10 Gbit/s upstream for back-up purposes and to keep latencies low between sites. PON technologies can't do this.  (Just to make it very clear, yes you can mix PON and P2P ethernet on a P2P network, just not on a low split network.
b. It doesn't exclude any future developments. There is no reason to assume Ethernet and PON will be the dominant technologies in 20 years time. Fibre is there for another 50 years. Any network design that doesn't make assumption over how people use it in 50 years is the best technology.

2. P2P allows the owner to raise penetration of the network better.
a. FTTH shouldn't be the only goal of a fibre network. The goal should be to get as many paying parties connected for as many different use cases as possible. Fibre lean is not a good idea then. If 70% of the cost of the network is in building the infrastructure, than the fibre bit is not the problem. Creating an option to hook up 5 times as more users than currently projected is not a stupid idea once you factor in mobile operators, municipalities etc as possible customers of the network. As the owner of the network, you shouldn't care about who connects as long as more people connect.
b. Multiple providers of wholesale services can operate on the same infrastructure. This is happening in Stockholm already. Stokab owns the fibres and if I have been informed correctly there are several operators of wholesale services selling to various ISP's and other service providers. For the owner of the physical infrastructure the question is not who sells the line, but if it's sold. (think of a huge office building.. it might be nice to have one tenant renting all the office space, but one tenant renting 50% of the space is not as good as 6 tenants renting 90% of the space.)

3. Strategic possibility to include or exclude certain operators
This one is most important for an incumbent trying to keep a position in every layer of the market. If a near monopoly in the infrastructure market is not good enough and a dominant position in the service layer can only be attained by nasty tricks and excluding other market parties, then low split fibre may be a good idea. Why make it easy for a competitor in the service market to enter the market. However if the infra investment is seen as separate from the service market, then opening up the infra might be a good idea. If you receive regula(r)(ted) returns on the infra, part of that money can be used to support the service business. Point to Point as said before allows a competitor to make use of the infra and still allow for competition.

Tuesday, 2 December 2008

Fighting over peering and transit

Scott Woolley , writing for , has written some excellent articles on peering and transit and the fights that Cogent got itself into in recent times. It offers great illustration of my explanation of Peering and Transit at Ars Technica.

Specifically I want to single out:
The day the web went dead - about the fight between Sprint and Cogent. Nice numbers are that Sprint started a fight over 0.004% of its revenue.
Metadata: Internet Innards - shows that peering is about power not morality
Telecom Knockout - On Cogent, its boss mr. Schaeffer and the fight with Telia

It really illustrates that the Peering and Transit market is very competitive and generally will come up with a good solution without governments intervening. (why do you think that most telco's don't want to see the voice business go that way? Telco's love government regulation. It has strenghtened the status quo in the voice business)

A great bonus but not peering and transit related is an overview of the money in the business:
Blown to Bits - on a digital world where all communications can be converted into generic bits and shipped over the internet for less. Here are the juiciest profit pools to drain (U.S. figures only). 

Thanks to Niels for the tip!

Monday, 1 December 2008

My internet connection was saved :-)

Thanks to the most highly valued reader who sent me the e-mail adress of BBNed's CEO. Thanks to the lady manning the phones at BBNed in Hoofddorp, who was kind enough to fire an e-mail to some people in the organisation at Friday afternoon. Thanks to BBNed's CEO, mr. Van den Ende, who forwarded my complaint and thanks to Bart at the Alice Helpdesk who was in contact twice today to tell me what was going on.

We're not going to be shut down :-) We might see a small interruption, but it shouldn't take too long.

Now I need to think again of something useful to post.