Optimising the NBN rollout

Mark Gregory’s impassioned opinions on a recent NBN-related presentation of mine deserve a swift reply. However, I would first ask that those genuinely interested in understanding what I was talking about to take the time (about 20 minutes) to listen to what I said, first hand, at the link noted above.

The reason for this is that the slide pack, on its own, doesn’t tell the full story here – instead it is merely the backdrop to what I said on the day. I have seen some comments critical to the content of the slide pack that are directly answered by listening to what I actually said.

I will now address the specific comments made by Gregory.

“Hackett argued that NBN Co should aim to cut features so that the Labor Fibre to the Premise (FTTP) NBN cost matches the proposed cost of the Coalition’s FTTN NBN.”

I argued no such thing. I argued that the NBN can achieve all of its intended consumer and business aims, at a lower deployment cost, by cutting out extraneous hardware and extraneous software feature sets that in fact add no benefit to the outcome for consumers.

First, to address Quality of Service (QoS),  QoS is only effective in networks with constrained bandwidth. However, the NBN has enough bandwidth. It is only the current NBN cost model that simulates bandwidth adversity in order to drive the price up artificially, which is NBN Co’s attempt to have consumers fund the current build cost.

But you don’t have to take my word for it.

There is an excellent treatise on this topic that was delivered by none other than Geoff Huston, the pioneer of the Australian Internet deployment (AARNet). Here’s Huston explain the reality of QoS in the modern world, via this video presentation from the RIPE65 conference:

To summarise Huston's talk:

·         QoS doesn’t fix anything, the damage just gets moved.

·         The amount of investment in deploying QOS is far more expensive than buying more bandwidth.

·         The carriers (in this context, NBN Co) are attempting to use regulation by government bodies to earn more revenue for no actual extra service.

Saving money on NTUs

Importantly, the existing NBN Co network termination units (NTU) can be deployed ‘by exception’ in the rare minority of sites where there is a site-specific need to have multiple independent, full-priced NBN connections at a single location. 

There are three key savings achieved by not delivering the NBN Co NTU (and battery – which only backs up the voice ports):

1.      The direct hardware cost saving of not deploying the box

2.      The direct labour cost saving of not doing a second site visit (“truck roll”) to consumer sites to install the NTU and battery

3.      The indirect reduction in network build cost obtained by moving to a multi-vendor approach in the rollout of the network in general – using competitive tension to ensure hardware supply costs keep reducing over time.

Conversations with NBN Co rollout contractors at the conference at which I spoke made it clear that in practice, two separate teams roll out the NBN connections in two separate passes. One team focuses on fibre deployment to each home or office, delivering the passive fibre port to the premises.

A second rollout team turns up later to install and deploy the active electronics, and finally a third electronics deployment happens when an RSP typically ships their customer a multifunction router (usually including voice ports and wireless features) as part of turning on a retail service.

Getting rid of the hardware drop

Gregory is incensed by my suggestion at the conference that NBN Co should dump three data ports and the two UNI-V voice ports from the NTD.

Indeed I went further than that – and proposed that we could remove the NBN Co hardware drop entirely, and allow the NBN Co fibre to be plugged straight into a GPON port on the customer router. For those rare instances where there is a benefit in that NBN Co device being installed it can be added in the future ‘by exception’.

There are two further issues raised in the article by Gregory that I should address.

Any assumption that I failed to do “identify what a next generation network should achieve and why”, as proposed by Gregory, is moot.  Such identification wasn’t the topic of my speech, so my not reiterating that well understood baseline is hardly a ‘failure’.

The implication that I would rather “have Australia spend $30 billion to swap out copper for fibre but leave everything else as it is” is entirely erroneous.

I would rather have no such thing. Nothing in the talk I delivered was about delivering a less effective result for consumers. Instead, it was (and is) all about delivering the same real world result for consumers at a lower cost and in a shorter rollout timeframe – a genuinely positive aim.

Gregory suggests that high bandwidth real-time services such as video conferencing, internet television, collaborative education and health services, require QoS and traffic class management. However (and I refer again to Geoff Huston’s talk), high bandwidth real-time services such as those noted by Gregory are all deliverable – and are routinely delivered around the world today – on fibre-based broadband networks without explicit QoS either added or required to be added.

The optimal price/performance solution on a modern internet-architecture network is always ‘more bandwidth’ in preference to ‘more complexity’ within the network core.

So this isn’t a case of achieving “fibre outcomes on a copper budgeting” by short changing the public on network speed or capability.

The cost savings are in hardware (deliver the NBN NTU and battery only by exception), rollout time and cost (remove one entire national ‘truck roll’ pass) and competitive tension (allow for multiple vendors in more of the end to end network). None of the things I have suggested as optimisations compromise the network speed in any way – indeed they arguably lead to the network running faster.

The truth about NBN Co ports

When it comes to the NBN Co ports, the purpose of the two phone ports should be as self-explanatory as it is ultimately un-necessary.

Gregory suggests that “Customers can take their existing handsets and plug them in which reduces the cost of purchasing network enabled Voice over Internet Protocol (VoIP) handsets”

The PSTN port in the NBN Co device is simply a VoIP Analogue Terminal Adaptor, and that can be supplied by a retail service provider where needed (either as a separate box or included as a built in feature in the customer router). There are other alternatives here as well – as discussed specifically in my talk.

The one thing that we both agree on is that most residential customers are likely to only use one port, because the other NTU ports are generally superfluous. Using two concurrent ports will cost double the monthly retail cost to a consumer – who is simply not going to do this when there is no advantage – only complexity – to be gained from doing so.

To suggest that telecommuters may use the second NBN Co data port to “connect to one RSP for work and another for home use,” is fanciful at best.

Telecommuters will do what they do today on existing broadband networks, and use a VPN client to obtain secure and authenticated access over the internet to their office. VPN technology is secure, established, and extremely convenient. VPN access can be established from computers, tablets, and/or automatically and transparently via an embedded VPN client inside most modern consumer routers.

Buying a second upstream internet connection for this purpose would double your monthly cost for precisely no improvement in outcome. No employer will fund that for staff when a VPN client is more flexible, secure and effective (and leverages the staff members existing internet link and costs).

CVC pricing a key consideration

In the real world, the data prices offered over the NBN will be absolutely dominated by the NBN Co CVC pricing – which is now the largest single cost component of the end to end data path from a consumer on the NBN through to their chosen data source.

There will be no ‘different data prices’ on the NBN as a result, until or unless the NBN wholesale cost model revises or deletes the CVC charge. Connections to business partners in the real world are achieved with VPN’s routinely or (in CBD environments) are also achieved routinely using far more cost effective connections from existing dark fibre backhaul providers.

Why would a business pay $20,000 per gigabit (plus GST) in CVC costs, per NBN tail circuit (hence $40,000 per gigabit link to connect two offices via the NBN), when they can pay a tiny fraction of that sum, today, for a dark fibre interconnect from one of the multiple competitive dark fibre business providers in the market today?

All a single business will gain from running two high capacity NBN Co internet services in parallel on a single site is a doubling of their underlying monthly access cost for no appreciable advantage.

The only rational use case for lighting up second and subsequent NBN Co NTU data points is where there are multiple absolutely independent businesses (or homes) in a single location. In those circumstances, then the NBN Co NTU can simply be delivered ‘by exception’ to that location.

The cloud computing furphy

The idea that cloud computing will be a major driver for more than one data port to be used, with businesses looking to “ reduce IT costs by connecting to RSPs that offer included free data when accessing cloud computing facilities,” is just a furphy.

There will be no such reduction in IT costs by connecting to RSPs offering free data to access cloud computing facilities. This is because no RSP is going to deploy a broadband NBN link to a business without passing on the underlying $20,000 GST per gigabit CVC cost to that business.

“Free data” only happens in retail broadband pricing models when the underlying data transport is at least 10x lower in cost than the current NBN Co CVC charge.

So what has NBN Co achieved by including four data ports and two telephone ports on the NTD? Gregory says it gives customers flexibility, opportunity and ultimately improved competition between RSPs.

On the contrary, what customers get (today) is an NBN rollout that costs more, takes longer, and is done with a single, locked-in underlying hardware vendor in the network core and edge, ‘forever’. Customers can churn between providers on a single port (without an NBN Co NTU) just as easily as they do so every day using existing ADSL2 broadband deployments.

There is no practical loss of flexibility for 99 per cent of consumers and businesses, and for the other one per cent, the NBN Co NTU can still be delivered (by exception) to offer multiple ports in a given location. Lower NBN Co rollout costs will be passed on to consumers in the form of lower retail costs, driven by lower NBN Co wholesale access pricing.

What the incoming government needs to do

The bottom line of my talk was to argue that there are a myriad of ways that the current NBN rollout could be optimised for both cost and rollout time, without compromising the technical outcome. 

The ideas I mentioned in my talk were merely examples, and I am absolutely confident that there will be massive opportunity for optimisation of the rollout to be obtained from an immediate and end-to-end audit of every cost and time component in the rollout of the NBN.

The single best thing that an incoming federal government could do for the NBN isn’t a cost/benefit analysis. Rather, it is that straight-up audit, followed by eliminating all avoidable costs that add nothing except rollout time and cost to the build. Experience in the internet industry over the last 30 years makes it clear that in all cases, when deploying a network, there is no substitute for very simple, very fast, very cost effective fibre-based network links. In all cases where QoS or other complexity beckons, it is always more cost effective to just turn up the speed.

And in the end, turning up the speed is precisely what the NBN is intended to do.

Simon Walter Hackett is the co-founder of Internode and a non-executive board member at iiNet.