National Broadband Network: Demystified

Well, this blog is finally back from a short hiatus.
Internode’s John Lindsay has posted some interesting details on the cost of bandwidth at various parts of the network (from peering and transit to last mile access).

These figures are the first we have seen since PPC1 went online in October. A previous presentation (PDF, 36MB) in June shows a big difference between then and now, with international transit costs (per megabit) falling $100, half the previous cost.

To summarize: Bandwidth costs, megabit per second

^ John Lindsay said:

It costs $110 a megabit to get the internet to me because we own the path all the way to the internet

I assume this figure includes the use of their own DSLAM (quoted as $10 per megabit per second in his June presentation). He also said:

But at the same exchange we have Telstra Wholesale ports and a customer connected to them costs us $200 a megabit

The cost of international transit has been subtracted accordingly. If we reverse this method with the previous international transit figure, we get $300, in line with the figure on slide 15 of the June presentation.

I hope these figures also put to rest any argument that one could viably run an ‘uncapped’ (without quota) internet service in the present climate.

edit: I got called on the last point, Internode released a truly ‘unlimited’ ADSL 1.5 plan on December the 8th. At $160 (without the bundle discount) obviously they are betting on heavy use. When we have unlimited plans for $50, that will be the day.

A lot has been happening in the past month related to the NBN, but I have been busy with more important things.

I promise there will be some new content in the near future!

This is something I have just whirlpool.net.au:

I figured its best to start off by finding out how much a ‘last generation’ broadband network cost to build:

Neighborhood Cable (now a subsidiary of TransACT) built a HFC network in Geelong, passing 50,000 homes, during 2002-2003.

At most they spent $31 million building the network, according to an ASX statements and annual reports during 2003.
(As of 31 Dec 04 their networks in Mildura, Ballarat and Geelong passing ~90,000 homes was valued at $47 million before they were bought out by their venture capital partners)

$40mil / 50,000 = $600 per ‘home’ passed.

This network is overhead-based. The active parts of the network (i.e node boxes) appear to be surplus Optus gear, which would reduce their costs.

They launched broadband services at $34.95 per month according to media releases at the time.

At most FTTH build costs $1000USD per home passed according to some statistics I aggregated last night. The real world numbers average around $800USD, which is ~$980AU. In particular, Verizon claims $700USD per home passed, and Embarq says FTTH build has reached parity with copper.
So, a monthly cost approaching $100AU does not seem too far off, but PON gear will no doubt fall in price further.. $60-70 is more likely.
This is just the cost to build the network

According to the BoS there are 7.4 million households in Australia, I am going to round this number way up to 10 for futureproofing and to help account some variation (i.e putting in place backhaul and underground works if demanded by NIMBYs)

BUILD COST OF NBN: 10 million * 980 = $9,800,000,000 (9 billion 800 million)

Verizon is said to have costed the FiOS rollout at $23 billion USD(source) to do at least 19 million homes.

In conclusion: The physical aspect of an NBN network is no where near $43billion. We already knew that, but some refuse to believe.

I am currently researching the numbers behind FTTP rollouts, in particular, ‘overbuilds’ where the FTTP network was built in the presence of a competing HFC (Cable) or existing copper network. These rollouts are also called ‘brownfields’ as the fiber can’t be stuck in the ground while the trenches are being dug.

In the mean time, I’ll publish links to useful information I find.

• Broadband Properties Summit 2008: Ingredients of a successful FTTH overbuild. Cites some examples of FTTH overbuilds in small to mid size towns.
• Verizon Policy blog: Why we are bullish on FiOS
• NY Times: Verizon’s FiOS – A smart bet or big mistake?
• Yankee Group: FTTH business model webinar (55 minutes) and slides
• Alcatel Lucent/Networks 2008: FTTX: Technologies and Economics
• Corning: The reality of FTTH in the United States (2005)
• Terraspan: Which method do you think will encourage brownfield FTTH
• FTTH Council US: FTTH Case Studies from three regional US providers
• Lightwave: Upgrade strategies for FTTH Networks
• Telephony Online: Embarq [US baby bell] slashes FTTP deployment costs, achieves party between copper and FTTP for new deployments
• Last Mile: Cost and profitability drivers for FTTH
• Lightwave: FTTH: ‘How much does it cost?’ vs ‘How much is it worth?’

It is heartening to see that there are many overbuild examples for small-mid size cities (i.e around the size of Newcastle or Geelong ) that have been successful.

nonbn.org was the catalyst for creating this site. Some of the claims made on that site are amusing, such as wireless broadband having unlimited quota. Currently, nonbn.org’s contention revolves around the sheer cost of the NBN. I am not an accountant or economist, so I will avoid analysis of the costings they have posted, for now. I will however, post links to articles about other FTTP rollouts and any figures they have posted as I can find them.

It is worth mentioning that claims, linked to by nonbn.org, by Professor Henry Ergas that an NBN connection will cost the consumer >$215 per month have already been shown to be deliberately misleading.

Enough of the economics for now.

The current state of wireless broadband, as retailed in Australia
The three mobile operators in Australia (Telstra, Optus and Vodafone Hutchison) all run 3G networks on the UMTS standard. This is the successor to the second-generation GSM standard first rolled out in Australia in the early 90’s.

GSM and UMTS are maintained by 3GPP (Third Generation Partnership Project), an international consortium. 3GPP is currently developing a successor standard to UMTS, called LTE, or Long Term Evolution, which has been in testing for some time now, with commercial launch overseas within the next two years.

UMTS, without addons, provides download and upload speeds of 3.84Mbit/s. Two upgrades to UMTS, HSDPA and HSUPA boost these speeds up to 42Mbit/s and 11.5Mbit/s in their most advanced forms. Currently, Telstra has deployed up to 21Mbit/s HSDPA on their NextG(TM) network, with the other carriers up to 7.2Mbit/s.

All these figures are theoretical! Real world performance is less!
42Mbit/s on the road sounds good, but you’ll never get it with current standards. Wireless is a shared medium, the bandwidth is shared with everyone connected to the base station, and additional overheads are present due to error correction, not to mention the signal strength between the base station and users.

Telstra themselves gives the following numbers:

• 550kbit/s – 8Mbit/s from a 21Mbit/s device, in metro areas
• 550kbit/s – 3Mbit/s from a 7.2Mbit/s device, in regional and outer metro areas, 6Mbit/s in CBD areas
• 550kbit/s – 1.5Mbit/s from a 3.6Mbit/s device, up to 3Mbit/s in some areas

Why the huge difference between the CBD and other areas you ask? While its plausible that base stations may need to receive equipment upgrades, there is a critical part of the wireless equation: backhaul, the link between the base station and the carrier network.

Historically mobile phone towers have been backhauled by older telco technologies such as multiple E1 links (n times 2Mbit/s), or in other cases, carrier grade wireless. The carriers are upgrading to conventional Ethernet to boost tower capacity. Upgrading every tower to Ethernet would involve fiber optic connections to each tower. And as usage increases, more towers will go up, all requiring fiber.

If we end up with say, a ~1km^2 radius for each network cell, it would be stupid not to let wireline users tap in. In fact, a future FTTP network would serve nicely as a mobile backhaul solution.

As an aside, wireless is not bulletproof. There are examples of 3G networks experiencing congestion and operators attempting congestion control by way of blocking applications.

Lets not forget the countless examples around the world of poor consumer satisfaction.

Enough of the current situation.

The horizon
Two technologies are in the running for the next generation of wireless broadband, LTE, as previously mentioned, provides download rates between 100-300Mbit/s, over 20MHz of spectrum. WiMAX, another technology, also has potential for fast speed, and is being rolled both here and overseas in various applications. (WiMAX comes in two versions, mobile and fixed; the coverage characteristics aren’t necessarily comparable).

While peak speeds of over 100Mbit/s sounds damn nice, it does come with strings attached.

The latest set of wireless technologies (LTE, and other radio based technologies such as the DVB family and DOCSIS 3) are approaching the limit for spectral efficiency, defined by the Shannon-Hartley Theorem. From now on, we only get higher speeds by throwing more radio spectrum at the problem. Radio spectrum does not grow on trees.

As an idea of peak spectrum efficiency, over 8MHz [Euro]DOCSIS can deliver 50Mbit/s of usable bandwidth, and DOCSIS 3 can bond 8 channels (64MHz) to deliver 400Mbit/s of usable bandwidth. This sort of efficiency can’t easily be achieved ‘over the air’, i.e 35.4Mbit/s per channel is quoted for DVB-T2 using the same QAM-256 modulation and the appropriate redundancies for free-to-air transmission.

If we do manage to free up that much spectrum over the air, it won’t last forever.

Lets not forget, thanks to ADSL2+ and cable, a good amount of Australian’s already get 10Mbit/s or more.

Trends in internet usage

Internet usage is rising, fast. According to ISP Internode, usage is rising 3% per month (source). Estimates in the US put traffic volume at 1 zettabyte (10¹⁵) by 2015. There is a exponential pattern to this, driven by:

• Increasing consumption of video – i.e from YouTube, and video on demand services like Netflix (US), Hulu, Blockbuster on TiVo (Australia). As screen sizes rise, so does the resolution of the video. 1080p will be old news in a few years.
• Changed habits as the younger generation, who have grown up on the internet, spend a lot of time on social networks, watch videos online rather than over traditional TV, and play multiplayer games
• ‘Cloud computing’ driving applications off home and office computers onto centralized servers in big data centers. In the long run it may be more efficient to return to the old days of computing and run intensive applications on grid services like Amazon EC2. People are also moving their data onto cloud storage services like Amazon S3 and Rackspace Cloud Files where the worry of hardware failure is taken out of the equation.

Preliminary conclusions
This has certainly been a long essay, and no doubt I will have more content on this topic. To sum up what I have said here:

• You need fiber for a mobile network. You will need lots of it to do the >100Mbit/s speeds promised in LTE
• Some 3G networks have barely survived as consumers sign up for broadband, what is to stop this happening in the future?
• To service users via wireless on the same level as FTTP we will need ridiculous amounts of spectrum in the future, to the point where FTTH will be needed
• Internet traffic volume will forever increase. Some commentators are uneducated about the adoption of technology by younger folk.

TODO: Explain why nonbn.org’s contention of wireless having unlimited quotas as being absurd.