Dave Taht wrote a blog post titled ‘My 2023 Predictions‘. It’s a thought-provoking piece about what might be around the corner and is definitely worth reading,
My expectation is that Internet operations in 2023 will see more of the same:
- More devices connecting.
- More use of higher-speed mobile forms like 5G.
- More growth overall (although from what Geoff Huston is uncovering in BGP, it’s possible to argue that the ‘rate of growth’ in BGP is slowing, and that’s an indication the dynamics of the global Internet are changing).
While forecasting is fun, it cannot predict the future. In 2022, there was one unpredictable outcome which, Dave notes, hit across the world pretty much at the same time, and has the potential to alter the surface of consumer and other Internet access. That’s Low Earth Orbit (LEO) Satellite Internet. Starlink has shown that there’s a high quality and fast option available ‘everywhere, all the time, all at once’ (within limits) if consumers want it.
Right now, my consumer (home) Internet is delivered as a fibre-to-the-home (FTTH, also called FTTP — P for premises) service. I’m well provisioned, at high speeds and low latencies — as low as one to two milliseconds for access to services in my metro area.
But only 30 minutes outside this metro area it’s possible to hit a speed wall and securing this class of service is difficult. There are some purely Australian reasons for this around the national network strategy, but it can be argued that they are largely irrelevant because across the world, and especially in the Asia Pacific region, the same problems can be seen for similar reasons. The cost of high-speed service delivery in the metro area is acceptable because of the density of customers for the investment in delivery. But move to ‘rural and remote’ and the cost shifts unavoidably against high-speed service.
Distance equals cost
Distance is cost — it’s the cost of fibre delivery, the cost of switching equipment, and the expense of maintenance. So, consumers can either move to cheaper forms of service delivery with lower capital expenditure, or slower forms of service delivery that can use the historical investment in telephony — the copper lines. That’s like walking back 10 years or more to the current cutting edge.
In many Asian economies with large rural populations, it’s a huge problem. On the list of infrastructure investment needs in developing economies, the Internet is high, but roads, hospitals, power generation, housing and sanitation are probably higher. And some of those (power for instance) are prerequisites to build out a fast network.
LEO Satellites are like their cousins Geosynchronous Earth Orbit (GEO) in their fundamental technology only. They are solar powered, they are launched on rockets, and they use very short wave (microwave and similar) radio spectrum to send high-speed signals. But the difference in orbit distance translates directly into a difference in time to send signals: GEO orbit satellites come with an unavoidable 500 milliseconds (half a second) round trip delay. An LEO satellite, on the other hand, orbits so close to the ground that it adds between 40 and 180 milliseconds of delay.
Back in the days of ADSL, 40 milliseconds was around the standard delay connecting to services in Australia. That’s one-tenth of the 500 millisecond-delay for GEO orbit. So, for voice and video, it removes nine-tenths of the visible delay. It’s a perfectly acceptable delay. And it’s remarkably consistent.
Measurement work done in APNIC Labs by Geoff Huston showed that Starlink can maintain the delay end-to-end, so while it may be higher than fibre domestic service delivery, it’s not variable. The ‘jitter’ (high variability in delay) component in IP use is not a problem. Jitter can cause significant problems for time-critical services like voice calls because the variability makes it hard to deal with consistently without exposing the delay variance to the users.
LEO means launching satellites. However, Starlink is using SpaceX reusable rockets to launch multiple LEO satellites at a time so the cost per satellite has been reduced. This is significant because, unlike GEO satellites, LEO satellites have to fly in ‘constellations’ to supply service to users. Consumers don’t connect to just one satellite; they move between them as they orbit. This represents another cost, but the cost of handling complex satellite receiving into a cloud of moving LEO satellites is amortized by Starlink into their Customer Premises Equipment (CPE) as a smart antenna with pre-calculated orbital information.
GEO satellites use larger dishes to avoid this cost, but they also tend to have less bandwidth to share over the same area. An LEO cluster of satellites can offer significantly higher bandwidth per customer, as long as the customers are far enough apart not to cluster onto the same satellite.
The cost of delivery of an LEO-based Internet isn’t as low as ground-based communications, which is currently being sold profitably for perhaps two-to-three times the cost of a metro high-speed service but it has comparable speeds and latencies. This is significantly better than a GEO-based service, which has broadly speaking, the same cost but for significantly slower service delivery on longer delay.
Just as 4G and 5G Internet offered the prospects of ‘cord cutting’ for users faced with either a speed bottleneck of copper wire service (ADSL) or the absence of any alternative due to high installations costs, LEO satellite services are offering rural and remote users a service comparable to metro fibre, without the high dependency on local distribution and reliable power that goes with it, making it a lower cost of deployment to reach the wider community.
Like the technology transitions from slow-speed modem to ADSL, from ADSL to high-speed cable and fibre, and simultaneously from 3G to 4G, and 5G mobile Internet, LEO is in some ways just ‘more of the same’. But it differs by representing a significant change in the cost of service delivery for a large footprint of users in the rural and less dense sector and with lower capital costs of infrastructure.
Just as emerging economies sidestepped the patterns of deployment in developed economies for mobile telephony and leapfrogged directly into 3G service delivery, perhaps there will be rapid uptake of LEO Internet services in those same economies, for reduced capital expenditure. This could mean that national infrastructure spend can focus on other issues like roads, health and power, which have concrete benefits to the community.
A possible unexpected side effect is that some traditional models of revenue collection through the former monopoly telcos, and content controls like Intellectual Property Rights management via IP geolocation, are going to have to adapt. The customer/consumer is still in the same place, but the agency of delivery has moved from an entity inside the economy to one outside. There are likely to be consequences for Internet governance not fully understood yet, as more swing from service delivery on the ground into LEO service delivery from an out-of-economy supplier.
Dave’s blog looks at far more than just LEO, and his core focus is on the bufferbloat issue that plagues older chipsets worldwide in computers and routing infrastructure. Dave notes that unfortunately the LEO systems are not immune to this problem, and work to engineer bufferbloat out of the network must continue.
Dave is right to flag the LEO as a moment of change in the Internet in 2023 and beyond. It’s going to have a very significant impact on how we deliver services to the ‘other 2.9 billion’, which is a significant goal of the worldwide communications community.
The views expressed by the authors of this blog are their own and do not necessarily reflect the views of APNIC. Please note a Code of Conduct applies to this blog.