One of the emerging debates over the NBN is about where the demarcation might occur between the parts of the network owned by NBNCo and the remainder of the market.
It's a debate that gets over-simplified in this characterisation: where will the "access" network connect to the "backhaul" network? In other words, where will the traffic that users generate on a service they buy from retailers - your phone calls and Internet access - connect to the rest of the world?
The relationships today
To understand how the NBN may change the structure of wholesale-retail relationships, let's take a simplified view of how networks connect together today.
Most commentary looks at both today's network and the NBN future as a two-tier wholesale-retail split. The NBN will provide a wholesale access network through which retailers will be able to sell high-speed services. There is, however, a third tier both in today's networks and in the NBN - the backhaul tier - whose role in the future is not yet clear.
Take broadband traffic on a current ADSL2+ service. It is delivered over a wholesale access network (Telstra's copper) back to a DSLAM in the Telstra exchange. That DSLAM may be owned either by Telstra (because you're a BigPond customer, or a customer of a BigPond reseller), or by someone else (iiNet, Optus, and any of the other eighteen DSLAM owners in Australia at the moment).
The DSLAM, however, is just an aggregation point. The "Internet backbone" connection doesn't happen in the Telstra exchange - rather, the traffic has to be transported from the exchange to the ISP's own network, because it's the ISP that buys the "transit" service connecting its customers to the rest of the Internet. To get from the Telstra exchange to its own network, the ISP also has to buy some kind of "backhaul" service - a high-capacity, aggregated link collecting traffic from the exchange and delivering it to the ISP's network.
Backhaul links have been the topic of much debate, angst, competition and animosity in the telecommunications industry for some time. In the cities, particularly to the east, the backhaul market is hotly contested by a large number of fibre owners; a smaller number of carriers, but enough for genuine competition, provide backhaul between the state capitals; but the further you travel away from those capitals, the less competition you will find. In most regional centres, the only company offering backhaul back to the cities is Telstra.
(This, by the way, is still an over-simplification: the wholesale-retail relationships that can exist in the broadband market can be Byzantine in their complexity).
Access versus backhaul
The NBN is posited as a wholesale access network. Retail ISPs will be able to buy capacity on the NBN to connect their services to their customers; how they decide to split up the 100 Mbps that the NBN will offer will be up to the retailer - some might want to offer 100 Mbps Internet access; others may choose to sequester some of that capacity for television services and offer a smaller pipe for Internet access. Just how the retail service-commercial models will stack up (and which ones will work in the market) remains to be seen.
What's missing, however, is a clear understanding of where the line will be drawn between that access network and the backhaul that will be required by anybody delivering services over the NBN.
To understand where backhaul connections might be made, we first need to understand the various technical structures the NBN might use - and for this, I turn to the draft deployment model recently published by the Communications Alliance, the Draft NBN Wholesale Service Definition Framework - Ethernet. It should be noted that this is a draft document only, but it provides a good understanding of how things might work.
The framework defines a more complex network architecture than most people realise. The framework recognises two possible access network architectures - GPON or direct fibre - both of which are aggregated in an EAS (Ethernet Access Service). The EAS forms the transport service that takes customer traffic back to the "service edge and core" part of the network.
The interface between the aggregation network and the core network might be presented as a "pure IP" wholesale service, or as an L2TP service (Layer 2 Transport Protocol), both of which, like the access network, are presumed to be wholesale services. These wholesale services deliver traffic to the retail service providers and thence to content providers.
Other options are imagined in the Communications Alliance framework. A service provider might, to some extent, be vertically integrated to some extent, and run its own IP or L2TP wholesale service without any third party, in which case it won't be a user of that third-party wholesale service. Another option might be called the "Google option" - the content owner doesn't' need a retail service provider, instead connecting "directly" to the core network.
Remember, however, that in the "Google option", the content will be hosted in some kind of data centre - the content owner will be buying its own fibre so as to connect from the data centre into the wholesale core network.
The NBN reference architecture
What we have in the NBN reference architecture is a model in which several wholesale layers exist: there's the wholesale access network, an aggregation / transport network (which may be thought of as fulfilling the backhaul role), and wholesale core networks.
So where might the backhaul come from in the NBN? There are only two possible answers to this question: backhaul could be provided by networks that already exist, or over a new network.
In the capital cities, a large number of backhaul networks already exist - but their structure has an interesting implication for how the NBN might be structured in the capitals. So let's take a quick look at the ways in which backhaul networks are now deployed within the metropolitan areas.
We have, of course, the extensive backhaul network deployed by Telstra, servicing its exchanges. Telstra and Optus also have backhaul servicing their HFC cable networks, from wherever the services are aggregated back into the core networks (I'm less familiar with the HFC architecture as it relates to backhaul).
In addition, Telstra and Optus have deployed fibre backhaul networks servicing mobile base stations, although not all of them: those base stations with fibre act as aggregation points, with microwave links serving other nearby base stations. However, the traffic demands being generated by 3G mobile broadband services are probably encouraging an expansion of fibre backhaul in the mobile networks.
Then there's the competitive fibre sector. Most of these focus on backhaul; for example, they might run between large data centres, and the "access" network is only that which exists within the data centres they serve. Or they might be built to provide alternative backhaul between Telstra exchanges and ISPs' facilities, in which case access and aggregation is the business of the ISP.
In a few cases, the new fibre networks almost completely ignore the way existing networks have been deployed. For example, a rail authority will use its own easements for its backhaul network, choosing the shortest route between its own aggregation points and a customer location, and only interact with other backhaul networks where it builds points of interconnect.
However, much of the backhaul that already exists within capital cities centres around the Telstra exchange, because the Telstra exchange is the locus of competitive broadband. Even though the NBN could bypass those exchanges, it still assumes that services will be aggregated somewhere between the fibre access network and the core network.
It may prove convenient for the NBN to mirror today's deployment geography to some extent - not because the network absolutely needs to use Telstra's exchanges, but because it might be the easiest way to ensure that customers can buy competitive backhaul without the competitive fibre owners having to drag backhaul fibre to new interconnect locations. That would put a premium on negotiation with Telstra, but negotiations are already taking place.
The missing link isn't in the cities, but the regions. Once you leave the metropolitan areas, backhaul becomes extremely limited. There are enough backhaul carriers to give us competition on inter-capital routes, but only a handful of regional centres have competitive backhaul.
This poses one of the more important challenges the NBN has to deal with. In a regional centre with no competitive backhaul, the retail service provider will be faced with the same problem as exists now: once its customer's traffic reaches whatever aggregation point exists in the centre, it needs a connection to the outside world - which, without inexpensive backhaul (whether priced according to regulation or competition), will mean a long and expensive trip to a capital city so that the traffic can enter the "core" network.
The market structure the NBN could create is, therefore, one in which at least three tiers of wholesale service (two of which are contestable) can exist. There's the NBNCo's monopoly wholesale access network; the contestable wholesale core network, and a contestable backhaul network. And this ignores other wholesale services that will act as inputs to the services offered to consumers - the buying and selling of telephony minutes, the Internet transit services, wholesale content and application services and the rest.
While the access network, the best-known and most visible part of the network, is getting all the attention at the moment, the cost of retail services offered over the NBN will be based on the interplay between all three of the wholesale network services a provider needs.