The ins and outs of international cables

Richard Chirgwin investigates the politics surrouding submarine cables, the business models on which they operate and their impact on Australian businesses.

When Pacnet CEO Bill Barney observed last week that Australia's Internet users would benefit greatly and quickly from investment in international links, he sparked an unintentional storm.

His remarks were taken as denigrating the NBN, because international links are so comparatively easy to build. In fact, his opinion wasn't news (he had made the same remarks in March during a press lunch in Australia), nor was he criticising the NBN (as he makes clear in a podcast interview with the author here, he supports the NBN but still believes that Australia needs more international capacity.

"Building new broadband networks is critical to the development of any country", he said. His prescriptions aren't restricted to the international sphere: Barney believes the development of broadband should embrace opening the last mile, and deployment of broadband as widely as is possible. Remote users, he said, become better connected to exactly those export markets they serve by having better broadband.

The remote user, he said - the farmer or the miner - is an export generator, and can get better linked into export markets by getting better access to broadband networks. The issue isn't whether those users need access to the networks, but how best to deploy and pay for their links.

International capacity is, as Barney believes, a fast way to build better links to the outside world. A new Australia-US cable would only take perhaps two years to deploy, compared to the much longer timeframes covered by the NBN.

But it is worth understanding the role that the submarine cables play in impacting both price and performance in Australia's broadband networks.

International Price Structures

When Southern Cross Cables announced a new price structure last week, it also published a chart indicating the historical price of international capacity, showing current capacity prices at around US 25 cents per gigabyte (around 30 cents Australian).

Price structures are more complex than that.

Providers generally buy according to capacity (for example, megabits or gigabits per second); prices vary according to how much capacity you buy (on a per-unit basis, it's cheaper to buy gigabits per second than megabits per second, if you can afford it); and different contract structures attract different pricing and different accounting.

However, at a national level, let's accept 30 cents per gigabyte as an accurate representation of the cost of getting content from America. There's the input into the ISP's pricing - with a few other things to consider.

The first is that the price cited by the Southern Cross Cable Network only refers to traffic traversing its own cable. At each end, the ISP also has to pay to get the traffic delivered to its own network (probably in an Internet exchange where Southern Cross terminates), and the Southern Cross Cable Network price doesn't include the upstream cost of Internet . And that's not to mention all the other cost components, like ULL/LSS access, rack space in the Telstra exchange, and so on.

So to get that 30 cents' worth of traffic across the Pacific, the ISP has to assume various other costs to pass on to its customers.

Be warned, then, that the rest of this article is going to ignore a number of other cost inputs so as to keep the international leg in focus.

What you get for 30 cents

For 30 cents, then, the ISP can buy a gigabyte of traffic. How far does that go?

Let's put together an ISP with just 10,000 customers signed up to an average plan of 1 Mbps.

Of course, the ISP doesn't buy 10,000 Mbps (10 Gbps) of upstream traffic to service those customers; it works out some kind of contention formula based on how congested its upstream links become. Assume that it finds congestion is manageable at a 50:1 contention ratio. At that level, its upstream service is about 200 Mbps.

If the links run at an average of about 75% congestion, the link is running at about 150 Mbps, and on these assumptions, the daily consumption is a little short of 13 gigabits, or 1.6 gigabytes.

Not all of the traffic will be headed overseas, however. If we assume an 80:20 traffic split, which will do for this exercise, then of the 1.6 gigabytes volume, the raw cost of international traffic is around $142,000 a year.

If those 10,000 customers are on an average plan price of $29 per month, the ISP's income is around $3.4 million, which means the ISP's international traffic (exclusive of tail costs and transit service) is about 4% of the total.

Remember that this isn't a rigorous analysis - it's more like a thought experiment. But there are a couple of points that undermine the calculation that international traffic costs about 4% of income.

The first is that at 10,000 customers, our ISP is too small to be buying international links directly from the submarine cable operator. A customer buying a single 155 Mbps service (known as an STM-1) will be on a higher rate than someone buying a 5 Gbps service.

So the small ISP will instead be buying its transit service in Australia, and paying a margin to its middleman.

The other issue for that ISP is that the customer base is changing. ADSL2+ is the fast-growth segment in the broadband market, to the extent that it's almost the default choice for new customers.

The way competition is working in this market, ADSL2+ services are on offer at the same price points as ADSL1 services. Although most customers live too far from their exchange to get 24 Mbps, they're getting service at much more than the 1 Mbps average I assumed in the first calculation.

What happens if the same 10,000 customers are running an average throughput of 5 Mbps?

If nothing else is changed, the raw international component of the ISP's cost base rises to around 20% of its income - a much more significant chunk of its cost base.

And, since our example is still a small ISP, it's not getting its international traffic at the best price available, or even the average price; its international downloads are marked up by whoever is offering the ISP its Internet transit service.

This assumed that nothing changed. In other words, the ISP can't raise prices (because that would put it out-of-step with its competitors), it maintained the same contention ratio so as not to upset its existing customers, and it does nothing at all to manage its traffic.

Since price is what customers notice most immediately, ISPs are more likely to adjust either their contention ratios, or implement various traffic management strategies, to cope with rising transit costs.

Capacity as well as cost

Apart from the cost argument, we need to remember that Australia's international capacity is limited: whatever is the sum of all the cables now in service (I don't have that figure to hand), the amount of traffic we can source from overseas, at whatever cost, is capped by the available cable capacity.

In other words, expanding the international capacity doesn't only reduce the price of international capacity, it also increases whatever infrastructure cap that might exist.

But it's most easily seen in the prices we pay; as any economist will tell you, if a resource is limited, and there's no external input to ration that resource, then it will be rationed either by queuing or by price.

And there is no doubt that such rationing seems to be happening. As Bill Barney noted, a 10 Gbps pipe between Australia and the US costs around $US30 million annually; the same capacity from Asia to the US costs around $US2 million annually; and between Europe and the US, he said he has seen prices as low as $US300,000 annually.

With such disparities, the real surprise is that Australia is able to maintain any kind of affordability in its services.

Footnote: Internet Transit

Internet transit is, to most people, the hidden part of their Internet service. We know that we can look at Websites hosted anywhere in the world, but we have only a vague idea of how the data travels onto our screens.

All ISPs have to pay for traffic going to and from the "backbone" providers (mostly in the US in the case of Australian traffic, with a growing proportion of traffic going to and from Asia). The cheapest place to buy Internet transit is America, but most smaller ISPs can't afford to set up and administer an international link, connect that link to a point-of-presence on the US West Coast, and operate equipment on the other side of the Pacific.

So they buy their Internet transit in the smaller, more convenient, but more expensive Australian market.

The international service is buried in the price of Internet transit bought in Australia, and forms a significant part of the price of services in Australia.





Read more on Data centre networking