What will the NBN look like (and who will be the winners and losers)?

Richard Chirgwin predicts the technologies and configurations likely to be deployed in Australia's National Broadband Network, and picks winners and losers once the project goes live.

Since there's only one topic on everybody's minds this week, let's talk the NBN.
The bare bones are already known to everybody: the government will establish a company to build the network, and fibre will be rolled out over eight years to “90 percent of Australians”. Users who are lucky enough to be on the network – and remember, there will still be two million or so of us who are still out of reach – will get a 100 Mbps connection to the home.
Investors, including retail investors, will be allowed to buy up to 49% of the network company during the first five years, after which the government will sell down its own stake (as a way of recovering its own investment), and since the new network will bypass Telstra's last-mile copper, it will also bypass the wrangling that many people blame for Australia's allegedly-lagging broadband adoption and relatively high prices.
At a recent press conference, Pacnet's CEO, noted that international links remain a huge influencer of end-user prices, and of course these international links aren't affected by the NBN announcement. Unless something dramatic happens to the price of international bandwidth, the new network won't have much impact on the cost of downloading content from overseas.
But there's the nutshell of the NBN. So now, let's dive into a couple of issues: what's the network likely to look like, and apart from Telstra, are there any likely “losers” when the network's rolled out?
What's on the Fibre?
Although the network hasn't been designed yet, industry talk favours PON – passive optical network – as the last-mile connection to the home. So let's have a quick PON primer.
No matter what technology you choose, it's inefficient to run a single fibre all the way from the exchange to the home. You can save on fibres, and therefore build cost, if you multiplex many user's connections onto a single higher-capacity connection somewhere between the exchange and the home.
That way, one fibre leaving the exchange can “fan out” to serve large numbers of end users. The issue for the network designer is to choose a technology that can do so cheaply and efficiently – and that's where PON comes in.
The key word is “passive”. There's going to be some kind of splitter between the exchange and the end user, and it's cheaper if you don't have to supply power to that splitter. If you have active electronics in the splitter, it will need not just power, but also cooling and protection from the environment.
In the PON, the splitter is a passive – i.e. unpowered – optical device.
It’s nearly Easter so we won’t try to offer a comprehensive PON primer lest we consume the entire long weekend, but there are two technologies currently in the frame: EPON (Ethernet PON), which falls into the IEEE standards stream, a highly-popular and well-established architecture; and GPON (Gigabit PON), which is part of the ITU standards stream and has been deployed by Verizon and trialled by various carriers worldwide. Both of these use different wavelengths to split downstream and upstream traffic.
It should be remembered, however, that the optic network itself doesn't define any services. It works at the lower layers; it will be up to the retail service providers to deal with the upper layers.
The network will, therefore, have some kind of provider aggregation point at the “exchange”. This will aggregate the customers signals to the different retailers selling services, and the retailers will take over traffic routing back to their own data centres at this point.
The new NBN will actually have three (at least) traffic aggregation layers: one in the optic network itself; a second, where the customer traffic is aggregated for the various service providers; and the third, which already exists, in which Internet traffic as aggregated to the backbone (ignoring aggregation for non-Internet traffic, and also ignoring the complexities of Internet transit and peering, a lengthy topic in their own right).
There is one other big consideration that's often overlooked in the enthusiasm for fibre-based networks. The NBN will shift responsibility for powering end-user equipment to the end users themselves.
At the moment, your PSTN phone (though not the cordless phone) is powered from the Telstra exchange, which feeds 48V down the copper pairs.
Telstra exchanges have various levels of power back-up: at least batteries, and in larger metro exchanges, diesel generators as well. So the exchange is able to keep the phones going during a blackout, either as long as the batteries have charge, or as long as the fuel lasts.
If the householder is supplying the power, the equipment will need local battery back-up so that the phones can survive a blackout. That's not a technical challenge, but because batteries don't last forever, it can be a formidable management challenge. Someone needs to be able to replace the batteries when need be.
Aggregating Traffic
Of considerable interest to the industry is the provider aggregation layer. The government has stated that the network will be open access, but what if the network architecture places cost barriers between providers and the network?
Just how access is achieved will be an important aspect of the network – not customer access, which is what the network is all about, but wholesale access to providers.
This hasn't yet been designed, so all we can do is speculate; but the physical layout of the wholesale side of the network – behind the “exchanges” – will be important.
If there's a very large number of aggregation points in which providers need a presence, then it will be more expensive to offer services, restricting national access to larger providers. Fewer aggregation points would be better.
Network design will, therefore, be of vital importance to the smaller players in the market. The largest carriers and providers have the money, expertise and personnel to cope with widely-distributed networks with dispersed equipment, but not the lower-tier players.
There is, therefore, a risk that in a country as large as Australia, the wrong architecture for the NBN would concentrate the market.
The upside for smaller players, if the network and its access arrangements are right, is that they'll have a national network and will be able to focus their attention on competing for customers.
While DSLAM-based ISPs are rightly critical of today's market structure, they do have an advantage over the mass of the country's 600 or more retail ISPs. Because they have the capital to build a network, they get better control over their cost structures – an advantage that will be diminished by the NBN.
Of course, the large ISPs will still enjoy considerable cost advantages in the rest of their network, but everybody will have roughly equal access to end users.
So there will be both winners and losers in the retail ISP market.
Will Local ISPs Lose Out?
Although the NBN proposal has been welcomed by infrastructure owners such as Internode (whose CEO, Simon Hackett, pointed out that the NBN build's timeframe leaves plenty of room to recover today's investment in DSL infrastructure while preparing for the new world of fibre), the NBN will affect the smaller, more localised infrastructure-based ISPs.
Australia has a number of regional providers who have invested in both DSL and wireless technologies. They generally focus on working at a local level and generating local loyalties, a world in which you know your customers by name.
These providers will retain the advantage of that loyalty, but they will find themselves under much stronger competition from the national sector which has hitherto left their territories alone as uneconomic. That segment of the market could go either way.
What About Indy VoIP?
There is one other market segment for which the NBN could represent a much greater threat: the independent retail VoIP market.
The VoIP market in Australia has stratified into two distinct segments: some services are offered by ISPs as a mechanism for customer stickiness; some VoIP services (including the ubiquitous Skype) are completely independent of any ISP, and accessible over any Internet connection (leaving aside those VoIP restrictions associated with cellular connections).
In the NBN, it's fair to say that “all voice will be VoIP”: carriers will almost-certainly aggregate their customers' traffic as IP, making IP telephony the natural choice for switching calls – although there will be a long period in which VoIP and PSTN services will still operate in parallel.
When all voice is VoIP, independent VoIP providers will find themselves losing their competitive advantage over the declining PSTN. At the same time, providers on the new NBN will be working hard to encourage bundling, because that will be the best way to craft packages that justify what's almost certain to be a more expensive connection than today's cheap-and-cheerful Internet packages.
Unless they can find enduring value-adds, independent VoIP providers may find it hard to compete on the NBN.

Read more on WAN performance and optimisation