Thursday, 11 April 2013

NBN: The Real Deal on New Network Technologies

I can't take any more of Turnbull's ducking and weaving on simple questions and his fatuous answers, especially on "What speeds and technologies will be available in the future?"

MT answers, "Oh, we just can't tell", either on the demand side ("what speeds will people want/demand/need?") or on the supply side ("what technologies will be available, at what price/performance?".

It simply isn't true. This is so disingenuous on his part as to be incredible.

We absolutely know, right now, what will be available commercially 10-15 years from now: just look in the labs!
This has been true since the days of the electronic valve and mid-1950's transistors.

The other Big Lie from the Coalition: "How long will this network last?" A: "Oh, we can't tell".

The real answer is: "The standard depreciation scale for this equipment is 20 years. We have to run it that long to realise it's full economic value".

Right now in Labs around the world we see a little work on wireless and pushing copper transmission, but the vast majority of work is on Optical Fibre:
  • Right now, you can buy commercial 40Gbps and 100Gbps SFP's (Fibre) for Ethernet.
    • The lasers & receivers needed to upgrade the 2.5Gbps used in GPON are already in volume production. GPON has a cheap forty-fold speed increase available before 2020.
  • CWDM and DWDM (Coarse and Dense Wave-Division-Multiplexing or "multi-colour" lasers) are in common service now for 1Gbps and 10Gbps Ethernet.
    • In 5 years, these will be available in volume production for GPON deployment. Cheap, proven technology, not fanciful.
    • CWDM gives 8-times the capacity over a single fibre without changing the infrastructure.
    • DWDM does 32-times or more, but at a higher price.
  • Multi-Terrabit systems, either single channel or massive WDM, have been demonstrated reliably in the lab.
    • These are likely to only be available as active Ethernet devices, for Carrier & Enterprise networks. But who'd have though 10Gbps would ever be available inside the home?
  • On the user side, answering "what good is all that bandwidth", there have been very large, long-running projects between multiple far-flung Universities over many years answering exactly that question:
    • The US had "Internet 2" project first more than a decade ago.
    • Australian Universities had a gigabit research network around the continent for most of the last decade. [more?]
    • Again, "look in the labs". There are well-known, good applications for high-speed networking to the premise.
      • What are now ordinary PC's and laptops will power quite extraordinary applications, given affordable high-bandwidth links.
  • The most focussed work on wireless is getting to Gigabit, a speed-point that Ethernet passed more than a decade ago and is now considered "slow" and obsolete in the enterprise.
    • There are two streams of development: mobile phone and WiFi, or 802.11 Ethernet.
    • There are two inherent problems and limitations with all wireless systems:
      • interference from "noise" like motors or lightning and from other active devices using the same band.
      • data over wireless is a "shared medium", like cable TV: per-channel, there is a fixed bandwidth available, no matter if there is one or a hundred devices sharing it.
        • Congestion destroys the performance of wireless services very quickly, unless they are "fixed", arranged to NOT share, point-to-point links.
For minimal upgrade costs, because it's a small electronics change at either end, we are looking at being able to upgrade speeds of the GPON NBN in 2025 by 320 times more likely 1,000 times.

If that sounds astounding to you, remember back to your 1998 internet PC: $2000, 400MHz CPU and 300MB drive, versus $800 now for a quad-core 3000Mhz PC with a measly 3,000,000MB drive. 40-400 times more CPU (it's not just Hz) and 100,000 times more disk. Moore's Law has been driving exponential growth of computing capability for over 5 decades. While we've hit some limits, like single-CPU "heat barrier", there is a lot more improvement coming.

Turnbull knows, or should know, these current fibre technologies and the upgrades in the pipeline.

He also knows that VDSL might be able to double speeds cheaply and because those unshielded, badly corroded old bits of copper in the ground are actually transmission lines for DSL, they obey the Laws of Physics. In this case, the Distance-Bandwidth Product: You can go twice as fast, but only over half the distance.

To increase the speed of the DSL NBN another factor of two, Turnbull has to halve the distance to the node. Which means deploying a whole bunch more nodes: this is hugely expensive.

That's the answer that Turnbull won't give: "We can get maybe 4 times more speed out of a DSL NBN, but at twice or more the cost, but for 2-5% upgrade we'll upgrades speed of a GPON NBN 300-1000 times, then the same again a decade after that if you want to".

If I were in charge of Telstra and wanted to take my business into the world of IP-everywhere, the converged (digital) communications we've been heading towards since the 1980's, would I want to keep copper with a very limited future, or embrace the only technology we know will scale-up simply and cheaply many hundreds of times?

Personally, I'd want to get into Fibre-to-Premises as quickly as I could to ride Moore's Law for as long as I could: More bits, more billable, better margins, higher profits, happy users, happy shareholders.

What we don't know, are two things:
  • What devices the technologies will turn up in.
    • Sony licensed the transistor from Bell Labs and surprised the US radio manufacturers who hadn't. In the US they built big, beautiful, expensive high-fi radios. Sony built a really small, really cheap ultra-low-fi handheld radio. Very quickly, every teenager had one.
    • Within 2 years, Sony had redefined the market and those Establishment companies were on the way out.
  • What the pricing of the production quantities will be. Determining consumer price-point is a "drop-dead" for all computer/communications products.
    • There have been so many "breakthrough technologies" in Computing over the last 60 years that worked in the Lab but have failed to make it into production because they couldn't be made, or if so, not economically.
    • The iPhone in 2007 took the long established "smartphone" market of Microsoft (they'd done it for a decade) and blew them away because they found the right price-point and minimum functionality.
      • This also blew away the previous game-changer, the Blackberry.
      • Which had blown away the previous, the Palm Pilot.
The only thing we know about the Future is: It will surprise us.

But we also know what made Apple great: The only way to predict the Future is to Invent it.

The Coalition is asking us to turn our backs on the only known, proven technology that can scale up cheaply thousands of times.
Once we spend all our money on a DSL NBN, the investment has an economic life of 20 years: it won't be "cost-effective" to change the NBN for 2 decades.

Two good papers from Alcatel, 2007, on GPON and VDSL:

No comments:

Post a Comment

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