Friday, 22 February 2013

NBN: MDU's, HFC and The Turnbull Plan

In the early 1980's I worked for a thoughtful, insightful Telecomms Engineer in O.T.C. Operations Branch, Brian C.

Brian made the heretical and, to me, counter-intuitive statement nearly 10 years prior to the first Analogue Mobile Phone service:
TV should be delivered over fixed-line and phones delivered over wireless with numbers assigned to people not places.
That's why today we have more mobile phone services than people in Australia and fixed-line telephone services have been in decline for 5 years or more.

We never got a universal Cable TV service, even in Metro areas, because of the Telstra/Optus dynamic, but we have had the Government shutdown analogue Free-to-Air TV to free up spectrum for better uses, like 4G phone and IP.

Thirty years ago Brian wasn't just right, he nailed it perfectly. So why weren't his insights accepted and acted upon by senior Telco management, even a decade or two later? If they had listened to these and other insights, their businesses, and profits, would've grown prodigiously, Australia would have world-leading Internet and mobile service and Telstra would also be a force in overseas markets.

The best technical & marketing insights were ignored because Telco's aren't managed to maximise profit to shareholders, utility to consumers or even implement Government Policy and National Priorities.

They are (solely?) managed to Maintain Our Birthright (read as Keep Our Monopoly at any cost) and Kill the Opposition!
Remind you of Microsoft, now in serious economic trouble and the opposite of the "way cool" it was in 2000?? Strikes me that way.

Prior to the NBN, the objective for Telstra was maintaining its (monopoly) copper Customer Access Network (CAN) as the only viable CAN. It had the economic size and market power to throw away a "few billion here, a few billion there" and to aggressively price products to force new entrants or existing competitors either out of business, or more usefully, just survive but not to thrive. That way Telstra maintains their effective monopoly and market/pricing power while technically not a monopoly or drawing the ire of the ACCC who can force some change, like ADSL1 pricing.

Brian's insight was based on two fundamentals:
  • Broadcast TV is unpersonalised, unidirectional, very wide bandwidth & to fixed destinations, and
  • voice calls are always person-to-person, bidirectional, low-bandwidth and "mobile" not fixed.
Radio, as a signal-superimposed-onto-a-carrier for free-air transmission, as opposed to baseband modulation carried on isolated/dedicated transmission media (cable), is perfect for many communications networks, public or private. AM & FM radio broadcasts cover large slabs of country cheaply and effectively, so car radios are now universal and ubiquitous.

Broadcast TV demands bandwidth, the more the better for extra content and services. Because everyone get the same service at the same time (the definition of 'broadcast') and TV sets have been large and fixed, fixed "pipes" have been a perfect match until now. With the advent of "small screens in every pocket", smartphones & tablets, fixed TV sets are now not so much of a given.

Broadcast radio, audio-only, is different in nature to broadcast TV and data (we could have Google, Murdoch & Fairfax broadcast Internet News), because you can listen and work/drive/walk, but anything that requires you to use your eyes must be exclusive. (SMS-while-you-drive is very dangerous).

The HFC network, remember we got only half a network each from the two Telcos, was built and provisioned for unidirectional broadcast: one signal to everyone. The definition of "broadcast"... It's a single shared "pipe" meant to provide identical streams to all customers.

BUT, the marketers prevailed and the engineers installed a low-bandwidth back-channel.
What were once very simple receive-only customer units, suddenly had to become much more complex and expensive transceivers, each capable of listening out for each other and not stepping on one anothers' toes. Plus, the essential internal network elements, uni-directional repeaters, had to be re-engineered to be bi-directional, more than doubling their complexity and radically increasing maintenance & troubleshooting problems. There is much worse also when you add the requirement "only registered devices can use the network".

Suddenly, engineers and network designers had to make another design decision and trade-off: where to allocate the available bandwidth? The downstream broadcast side versus the bi-directional data path?

Where did the revenue come from? Content Providers, at least customers tuning in to their broadcast TV services.
So that's where the engineers allocated the majority of the available bandwidth.
The Data Path, especially the shared upstream connection, started as a "whopping" 10Mbps. If you're the only kid on the block using it, latency and throughput are respectable, even now.

Because we have remarkably synchronised lives in the Western/industrialised world, all public Comms services have "Busy Hour", or times of Peak Demand (e.g. 6PM-10PM), that the network must be dimensioned for... The Peak Demand can be easily 10 times more than average daytime demand, and overnight it shrinks to "nearly nothing" as we all sleep.

The User Experience is what users see during Peak Demand periods (latency and throughput), not the nominal connection or single-user speed, nor the attainable rate aggregated over 24 hours.
The ACCC might well insert itself into this arena if HFC services are advertised as "high-speed". That could be a major problem for the Turnbull-Fletcher-Abbott NBN.

HFC network owners know all this. Why don't they just redistribute the available bandwidth, cut back on TV channels and dedicated them to Data Paths?
Because every single active device, and many/all passive filters, would have to be replaced or upgraded. This "visit every device" scenario is around the same cost as installing a new network, much more if done in a piecemeal/ad-hoc fashion.

AND, like the cutover to Digital TV, it's a step-change. Existing services stop dead...
Plus, you get very limited extra capacity for your money. For the price of a full-fibre upgrade, you get one tenth the Data Path capacity to sell.

Any sane HFC network owner presented with this economic proposition will do exactly what Optus and Telstra have done: Milk the Cash Cow for as long as it can with minimal maintenance, and wait till someone provides them with a free upgrade to full-fibre...

The arguments against basing a fast broadband service on HFC infrastructure are:
  • Coaxial Cable is a shared infrastructure, it doesn't scale past a handful of subscribers.
    • Internet traffic is bidirectional, point-to-point.
    • HFC networks are designed for unidirectional broadcast. It's the worst fit possible.
  • The Shared Everything model is the worst of all worlds.
    • complex protocols with really poor error recovery
    • congestion or fault equipment can lead to zero throughput.
    • maximum speeds are limited and then shared.
  • Coaxial Cable, set-top boxes and transceivers are "so 80's" technology:
    • expensive
    • low-volume production runs and very old technology components/designs.
    • power-hungry
    • low-bandwidth compared to Fibre
    • End of Life/Fully Mature:
      • no expansion/upgrades possible
  • Shared Bus network are terrible in operation.
    • Single equipment failures take out the whole network and like wired-in-series Christmas lights are painfully slow (expensive) to diagnose.
    • Long-distance telecomms ditched coax cable on land and undersea in the 1980's in favour of fibre: faster, cheaper, upgradeable and high quality over longer distance.
    • Ethernet only took off when 10-base-T, with switches and dedicated twisted-pairs per device, not shared coax and bridges, came along.
    • The initial LAN competitor to Ethernet, IBM's Token Ring, also failed because of limited throughput due to single shared resources.
    • The original Ethernet, the "Aloha" radio network around the Hawaiian Islands failed due to signal-noise, congestion and faulty transceiver blocking everyone.
  • Even in the most favourable market, Internet-over-Cable hasn't dominated.  In the highly competitive and relatively unregulated US market which has multiple successful, large Cable TV companies, Cable/HFC hasn't taken over. If it really had potential, it would be "King" somewhere, but it isn't.
Compared to nothing, Cable-Internet is terrific.
At scale, it's complex, expensive, slow, failure-prone and power-hungry compared even to ADSL1.

Recent NBN/HFC stories, including mention of Multi-Dwelling Unit (MDU) problems.

David Braue confirms that HFC-Data has highly variable performance, depending on demand, often becoming unusable in regular periods of Peak Demand.

Why does wiring-up subscribers, for either NBN-fibre or HFC, in apartments blocks or town-houses (MDU's) present any problem in Australia?

We know it does because even after nearly 20 years of HFC networks in Australia, MDU's are still not connected.
Contrast this with Transact's VDSL1-FTTN installed a decade ago including into MDU's. Running "cat-5" cable alongside phone lines into buildings and up the internal risers was simple and cheap - and like phones, just one cable could be economically run into a building and not require any special "Body Corporate" access, costs or permissions.

Renai LeMay addresses exactly this issue:
The HFC cables of both Telstra and Optus runs right down this street; right past my apartment. However, for the past decade that I have lived here, I have been completely unable to get this infrastructure connected to the two apartments in which I have lived, because to start with, the apartment owner of the entire block won’t get it connected to every apartment as both telcos require, and so I can’t get it connected to my individual apartment. I’ve asked several times.
For both HFC-cable and NBN-fibre, there are more complex installation requirements and importantly, common-infrastructure required in Body Corporate locations, the costs to be shared by all owners, connected or not.
For an individual occupier to get connected from the common-infrastructure, someone has to pay for the install. In rented properties, owners aren't interested in paying for something they'll not get utility from or increased rent, and renters, often the majority of occupants are disinclined to pay non-refundable install fees.

The short answer to the question, "Why in Australia are MDU installations of Cable and Fibre so rare?", because you have to get a whole bunch of people to agree, and in every building.

If you've gone to any Body Corporate meetings, even of small buildings, getting anything done is a nightmare, let alone getting something through that under half the residents will use.

Unless the install is free and can't be easily refused, most Body Corporates won't spend the money.

How did this get overcome in the USA?
I presume different laws for real-estate and Telecomms.

No comments:

Post a Comment

Note: only a member of this blog may post a comment.