analysis Will cheaper ISPs provide a degraded level of service on the NBN compared to ‘premium’ ISPs, through the use of poorer contention ratios? We’ll look at both sides of the issue in this follow-up article on the future of retail ISP competition under the NBN.
On Thursday I published an opinion piece exploring what retail competition between Australia’s ISPs would actually look like in the new world of the National Broadband Network which is gradually dawning. In this article, I argued that competition on the NBN would rest almost solely on price, as the importance of other differentiating factors between ISPs like Telstra, Optus, TPG and iiNet would diminish to zero as the network was rolled out.
As the discussion following the article progressed, there were a number of different responses. Some readers agreed with me, while others emphasised what they saw as the continuing importance of factors such as customer service levels, the availability of content from different ISPs, and other value-added services. However, as the debate continued, one issue in particular began to stand out as a controversial one that would not go away: The issue of network performance.
One of the main differentiating factors between the different ISPs offering retail NBN broadband services, many readers argued, would be the extent to which they provisioned adequate backhaul bandwidth to service customers’ connections. In short, every NBN customer should theoretically get the same speeds on the fibre running out to their premise, but real-world network speeds would depend on what bandwidth their ISP had provisioned beyond the fibre access network.
Now, this debate is very old one, and my own opinion on it has changed several times. With this in mind, in this article I want to go into the recent history of this debate, and then look at the current contention ratio discussion with respect to the NBN. I’ll weigh the evidence on both sides and try and come to some conclusions which may help further the discussion.
Firstly, it’s important to note that the contention ratio issue does exist with Australia’s network infrastructure today. In fact, it is one of the core issues in play when we talk about how well the various broadband networks around the country perform. As a number of readers have highlighted, some ISPs, notably TPG, do currently suffer from a large number of complaints on their forums on sites like Whirlpool, with users criticising the company for slow ADSL broadband speeds, particularly at night when more home users are more fully utilising their home broadband connection.
Whether it’s “speed issues”, “packet loss”, “slow download speeds”, “congestion” or as one user put it, “dropout-ageddon”, it seems clear that TPG continues to receive more complaints about this kind of problem than other ISPs do, leading to long-running speculation that the company is not provisioning enough bandwidth to each telephone exchange to be able to support the number of customers connected. At night, so the highly speculative theory goes, TPG’s network gets overloaded and customers can’t get the high-speed broadband they’ve paid for.
The issue of contention ratios is also one which affects other forms of network infrastructure in Australia. If you dig to the bottom of problems such as Optus’ network hiccups after it won a huge slice of the iPhone market in 2008, Vodafone’s ‘Vodafail’ problems in 2010 and 2011, or even the sporadic complaints from those on the HFC cable networks operated by Telstra and Optus, you’ll find that the problem is fundamentally the same; too many users, on network infrastructure that didn’t have enough capacity provisioned to handle their download demands.
So will this congestion issue apply to the NBN, particularly the NBN’s fibre rollout?
In the past, I have argued that it will. In late July 2011, I published an article arguing, much as many of Delimiter’s readers have argued over the past several days, that the network topography of an ISP’s connection to the NBN featured many areas where an ISP could skimp on the services it provided to customers by under-provisioning bandwidth, similarly to the way they currently can with Australia’s copper network. In fact, much of ISPs’ networks won’t actually change much, when the NBN is introduced — in some cases, just the customer access part of it.
In short, you replace copper with fibre, but congestion can still occur in the rest of the network. I recommend you check out the excellent diagrams included with that article, as supplied by networking engineer Mark Newton. At the time, this was a popular argument, and it was backed by the opinion of experts such as then-Internode MD Simon Hackett, who also argued that each ISP would have a different contention ratio which they would apply to the numbers of customers supported by each parcel of backhaul bandwidth at each NBN point of interconnect.
Now, some elements of this argument are still valid. However, since July 2011, a number of other factors have actually changed, leading me to believe that there will be much more similarity between the performance of NBN ISPs than was previously believed. I’ll go into why I believe that now.
When we last discussed this issue in depth in July 2011, Hackett disclosed that tests of the NBN’s fibre in Tasmania had shown that when ISPs provisioned backhaul to users, they would need to apply at least 200Mbps of capacity to each NBN point of interconnection, in order to be able to guarantee the NBN’s 100Mbps speeds to each customer.
Why 200Mbps? Said Hackett at the time: “Simply, so that more than one customer at once is capable of achieving the advertised 100 megabit download speeds the fibre can provide. This results in a need to have 100 megabits to allow any 100 megabit service to work, plus another 100 megabits of ‘burst’ capacity (for a total of 200 megabits) to allow for the transient needs of more than one customer using the network at once.” 200Mbps could actually serve “quite a lot of customers”, according to Hackett; although eventually, of course, ISPs would need more so-called supporting ‘Connectivity Virtual Circuit’ (CVC) bandwidth as they added on bulk customers to their networks.
At the time, it did indeed appear that not every ISP would want to pony up for a full 200Mbps of CVC connection each month, given that this would cost $4,000 per NBN PoI per month; meaning that if you wanted to supply NBN services nationally to the network’s 121 PoIs, you’d be paying almost half a million dollars per month. And the issue of contention immediately raised its ugly head. I’m sure that the number-crunchers inside ISPs were giving themselves a great deal of headaches at that point as they tried to figure out how the NBN profit equation could work, especially in the network’s early days, and how close to the line they could push network performance.
However, several things have changed since this time.
Firstly, in August 2011, NBN Co announced that it would offer customers a rebate on the first 150Mbps of CVC pricing to ISPs until there were 30,000 premises passed in what it described as “a connectivity serving area” — which connects to one of its planned 121 points of interconnect. Given that each ISP will only be taking a slice of the total NBN fibre market, what this means is that in the short to medium time frame (that is, much of the next decade, for many areas), it will be quite hard for contention ratios have a big impact, given that the substantial rebate on CVC pricing would suggest each major ISP will be able to afford to adequately provision their new NBN customers.
In the long-term, another factor also comes into play.
Since mid-2011, Australia’s ISP industry has gone through substantial consolidation. In November and December 2011, iiNet bought substantial slices of the Canberra and South Australian markets through the purchases of TransACT and Internode, and in April M2 bought Primus.
These moves, along with previous acquisitions, usually by iiNet, have had the impact of basically consolidating Australia’s fixed broadband market into just four major players — Telstra, Optus, iiNet and TPG (keeping in mind its acquisition of PIPE Networks), all of which now have very strong networks, including extensive fibre infrastructure, access to multiple redundant international fibre connections, and deep pockets of capital to soak up costs. A couple of other substantial players — principally M2 and Dodo — are bringing up the behind.
Now, with the previous generation of smaller ISPs, I would find it easy to believe that they would find it hard to be able to afford to provision enough bandwidth to each NBN PoI to meet their customer requirements while making some profit on top. In fact, this very issue was raised by Internode’s Hackett in July 2011; and Hackett cited an inability for Internode to be able to compete in an NBN world as one reason why he sold the company to iiNet.
However, despite the complaints about a perceived level of congestion on TPG’s network, I find it hard to believe that the company — or any of the other three major players — will not provision enough bandwidth on the NBN to meet their customer commitments in each area, especially given the 150Mbps CVC rebate head start they will have. All of the four players left have more than enough scale to get this job done and take some profit, without resorting to such levels of degraded service. We don’t see that kind of behaviour from Telstra, Optus or iiNet today, and it’s hard to believe that much of it will be seen from TPG.
As then-Exetel chief executive John Linton wrote in July 2011, if NBN ISPs do under-provision bandwidth, it would be “obvious”, given that the NBN’s underlying fibre technology is so inherently reliable and stable. If TPG or anyone else starts skimping, tests will very likely be able to determine this. And also bear in mind that churning to another ISP on the NBN can be done virtually instantly; there should be no fortnight-long gap without broadband to disincentivise people to switch ISPs.
Now, I’m not going to say that congestion won’t be an issue at all on the NBN’s fibre; certainly I think the issue of network congestion is one that has existed as long as humanity has built communications networks of any kind, and it will continue to exist until we greatly surpass our society’s current knowledge of theoretical and applied physics.
I think that when it comes to the smaller crop of ISPs in Australia, particularly those buying NBN access through a wholesaler and not directly from NBN Co, congestion will be a topic of much discussion amongst their users. And there still remains the possibility that low-cost ISPs such as TPG will take a ‘closer to the line’ approach than more ‘premium’ ISPs such as iiNet.
For example, I would consider it reasonable to believe that a lower cost ISP would provision only 175Mbps in a scenario that an ISP like iiNet might provision 200Mbps. Or alternatively, to pick an example out of the air, they might allocate 200Mbps per 200 customer premises, where a premium ISP would provision 200Mbps per 175. I don’t know the precise ratios, but I’m betting that a low-cost ISP would be watching the performance of every NBN PoI closely and pushing things closer to the line than a ‘safe’ ISP like iiNet.
There is certainly evidence that ISPs are currently looking in detail at this area as they try to weigh up what their financial models will look like in an NBN world. In September, for example, Primus general manager of marketing and products Andrew Sims declared his company was “pretty comfortable” with his company’s NBN pricing at that time, noting that the ISP wasn’t targeting “the higher users” in the market, and that there was still a bit of work to do to examine specifically how contention ratios in the network would change as more users started to adopt the NBN.
However, a further factor also applies here. Let’s assume that some cut-rate ISPs do under-provision bandwidth on the NBN. One has to wonder whether to what extent that the majority of ISPs’ users would even notice. As Shadow Communications Minister Malcolm Turnbull has pointed out a number of times (and I agree), there are very few applications which will actually consistently require the full 100Mbps (eventually 1gbps) speeds which the NBN will offer customers. Even high-definition video streams will only require something in the order of below 10Mbps each at most.
When Internet users are using an ADSL broadband network, which is limited to 24Mps in theory and usually below 16Mbps in practice, any speed impact on their connection is readily apparent. But when much faster, lower latency fibre connections are in widespread use, congestion would need to be heavy indeed for most users to notice there is a problem.
For example: Is your 100Mbps connection only delivering real-world speeds of 75Mbps? Or even 50Mbps? No problem: That’s still way more than enough bandwidth to deliver the speeds you need to do almost anything, in a very timely manner, on the Internet, compared to the paltry ADSL2+ speeds most Australians use today. Under-provisioning by only a small amount could have a drastic impact on ADSL congestion. But given that the NBN’s extravagantly high speeds won’t usually be tapped to their full extent, I think ISPs will be able to get away with a lot more when it comes to the network’s fibre.
Delimiter reader Dan also raised another tantalising possibility this week: What about if NBN ISPs stopped guaranteeing maximum speeds available under the NBN and started guaranteeing minimum speeds? For example, what if ISPs started playing with contention ratios as a commercial offering, offering a series of plans like this:
- 100/40Mbps maximum, 12Mbps minimum, 500GB data, $300/month
- 100/40Mbps maximum, 4Mbps minimum, 500GB data, $160/month
- 100/40Mbps maximum, 1Mbps minimum, 500GB data, $80/month
In this scenario, the first plan would have a much better contention ratio than the second plan, which would in turn feature a better ratio than the third plan. Users would have the guarantee of minimum speeds under all, and the likelihood of regularly bursting to much higher speeds, but the guaranteed minimums would change, depending on how much you wanted to pay per month.
Are these examples where NBN contention ratios would apply in a good manner, or a bad banner? I think it’s an interesting example where much of the choice would actually be left up to the user. Just how good a service are you prepared to pay for? One suspects that NBN Co’s boffins have considered these kinds of scenarios at some level, but that few others in the ISP industry have.
Overall, what it feels like to me here is that with respect to the NBN, both the commercial and technical aspects of the network currently appear set up in a way that will limit any congestion problems for the vast, overwhelming majority of Australian broadband users in the vast majority of cases; especially when compared with today’s broadband networks.
Of course, I also feel that the dynamics of this situation have changed a great deal over the past 12 months; and they could change again. Perhaps in 12 months’ time I will need to write another analysis piece going into the situation again. But then, that would be an endeavour which I would enjoy greatly, so it’s all OK ;)