Monday, December 30, 2019

Starlink simulation shows low latency without inter-satellite laser links

Handley's simulation shows that, while not as fast as an equivalent ISLL path, long bent-pipe paths would typically have lower latency than terrestrial fiber routes between the same two points.

Mark Handley, a professor at University College London, has made two terrific videos based on runs of his simulation of the first -- 1,584 satellite -- phase of SpaceX's Starlink Internet-service constellation. I discussed the first video, which assumes that the satellites have inter-satellite laser links (ISLLs), in recent post.

While SpaceX plans to deploy ISLLs in the future, their early satellites do not have them since at 27,000 km/hr they are state-of-the-art technology and may also encounter political problems in some nations. Since it could be a year or so before SpaceX begins launching ISLL-equipped satellites, Handley has made a second video that assumes the phase one satellites do not have ISLLs. This post discusses that video.

Satellite footprint (source)
Each satellite has four phased-array antennas that can rapidly switch narrowly focused connections to terrestrial antennas falling within a large "footprint" area. The terrestrial antennas might be Internet-connected ground stations or end-user terminals. If there were no ISLLs, long-distance traffic would have to be relayed by bouncing packets up and down between satellites and the ground.

Many people -- me included -- have assumed that these "bent pipe" hops would significantly increase latency on long-distance paths, but Handley's simulation shows that, while not as fast as an equivalent ISLL path, long bent-pipe paths would typically have lower latency than terrestrial fiber routes between the same two points.

Sample Seattle-New York path
Consider, for example, this six-hop route between Seattle and New York. The bent-pipe route has a round-trip time of 36ms versus 78ms for the current Internet and 38ms for an hypothetical great circle fiber route, which would be impossible because of mountains and other obstructions.

That example was taken from a run in which only six orbital planes had been populated and it assumed ground stations at popular SpaceX locations plus a few others that Handley assumed would be added at strategic points. With only six orbital planes, global coverage is spotty but it is solid around 53 degrees north (and south). As more satellites are added, coverage becomes wider and latency improves. By the time all 1,584 satellites are in operation, there is global coverage and latency is consistently better than today's terrestrial Internet.

User terminals as relays
While adding satellites improves performance, adding additional ground stations has an even greater impact. That suggests the possibility of relaying traffic through idle end-user terminals, which also have phased array antennas. Handley ran a simulation assuming relays every 100 km and found that latency across the US was roughly cut in half and jitter (latency variance) also declined, but the number of route changes increased to about one every five seconds. That sounds like a lot of overhead but Handley feels that it is feasible to handle. It would also require a more expensive user terminal, a little power and the permission of the user so SpaceX might subsidize the terminals or charge less for service.

Handley also considered inter-continental relaying, which would require relay stations on strategically placed ships at sea. (It turns out that no ships would be needed to cross the Northern Pacific, but that would require a relay station in Russia, which might be a political problem). He doesn't mention the possibility, but couldn't cargo and cruise ships act as slowly moving relay stations? (They will certainly want to be terminal-users).

The example shown above is for an east-west link but Handley also looked at long north-south links and found that ground relays actually beat ISLLs in some cases and were always better than fiber, but the best results are achieved by a combination of ISLL and terrestrial links, which we can look forward to once SpaceX and others begin deploying satellites with ISLLs.

Handley concludes by pointing out that since he started making the video, SpaceX had revised their constellation configuration from 24 66-satellite planes to 72 22-satellite planes. It turns out that once the first 1,584-satellite phase is complete, there is pretty much no difference between the new and old configurations, but it does require a few more satellites to be deployed before the trans-Atlantic and Pacific relays will work continuously. Note that SpaceX hopes to complete the first phase by early 2021.

I can't conclude this post without mentioning Handley's charming disclaimer that he has no inside information, but, based on public statements, has made reasonable assumptions about "what they could do if they wanted to, but probably isn't what they will actually do."

Watch the video:



Update 1/1/2020

Handley presented a paper on this research at the 2019 ACM HotNets Conference. You can see a video of his presentation and download a copy of his paper here. (The video of his talk is free, but the paper is behind a paywall).

Update 1/13/2020

There has been further discussion of this topic in the Reddit Starlink Community. Commenters have pointed out that ISLLs are cutting edge technology and the current cost of Mynaric's 10 Gbps terminals is prohibitive, though it will doubtless drop with mass production (and Mynaric has hired an ex-SpaceX executive). While SpaceX has announced plans to launch satellites with ISLLs by the end of 2020, those may just be for testing. Furthermore, they will have many legacy satellites in orbit by then and those will not be replaced for around five years. SpaceX will not have a 100% ISLL constellation until 2026. Perhaps OneWeb made a wise decision in postponing ISLLs and Wall Street arbitrage traders will have to wait a few years for ultra-low latencies.

Sunday, November 03, 2019

What to expect from SpaceX Starlink broadband service next year and beyond

Last May, SpaceX founder Elon Musk tweeted "6 more launches of 60 sats needed for minor coverage, 12 for moderate" and SpaceX President and CEO Gwynne Shotwell recently said they planned to be offering service in parts of the US in mid-2020, which would require six to eight 60-satellite launches. The first of those launches will be in the middle of this month on a thrice-flown Falcon 9 booster. (They will also need customer terminals and Elon Musk has used a prototype to post a tweet from his home).

Six to eight launches would bring them up to Musk's "minor" coverage by mid-2020 and, if they maintain the same launch rate, they would achieve "moderate" coverage around the end of the year. But, what is meant by "minor" and "moderate" coverage? A simulation by Mark Handley, a professor at University College London, provides an approximation of the answer.

The first Starlink "shell" will have 24 orbital planes. Each orbital plane will have 66 satellites at an inclination of 53 degrees and an altitude of 550 km. Handley ran simulations of the first 6 and first 12 orbital planes -- corresponding roughly to the SpaceX plan for 2020. Snapshots of the coverage area "footprints" from the two simulations are shown below:

Coverage with six and twelve 66-satellite orbital planes

The blue areas -- around 50 degrees north and south latitude -- are regions with continuous 24-hour coverage by at least one satellite. With six orbital planes, there will be continuous connectivity in the northern US and Canada and much of western Europe and Russia, but only southern Patagonia and the South Island of New Zealand in the sparsely populated south. Note that the financial centers of London and (just barely) New York will have continuous coverage, but, since these early satellites will not have inter-satellite laser links (ISLLs), SpaceX would have to route traffic between them through an undersea cable.

Coverage is continuous around 50 degrees north and south.

(At this point, you should stop reading and watch the video (6m 36s) of the simulation which shows the footprints moving across the surface of the planet as it rotates).

With 12 orbital planes, all of the continental US and most of Europe, the Middle East, China, Japan, and Korea will be covered. Shotwell says that once they have 1,200 satellites in orbit, they will have global coverage (with the exception of the polar regions) and capacity will be added as they complete the 550 km shell with 1,584 satellites. That should occur well before the end of 2021 since she expects to achieve a launch cadence of 60 satellites every other week.

Shotwell also said they planned to include ISLLs by late 2020, implying that less than half of the satellites in this first shell will have them. Those ISSLs will give SpaceX an advantage over terrestrial carriers for low-latency long-distance links, a market Musk hopes to dominate. ISLLs will also reduce the need for ground stations. (Maybe they can lease ground-station service from SpaceX competitor Amazon in the interim)

All of this is cool, but what will it cost the user?

it sounds like SpaceX is serious about pursuing the consumer market from the start. When asked about price recently, Shotwell said millions of people in the U. S. pay $80 per month to get “crappy service.” She did not commit to a price, but homes, schools, community centers, etc. with crappy service would pay that for good service, not to mention those with no service. Some customers may pay around $80 per month, but the price at a given location will be a function of SpaceX capacity, the price/demand curve for Intenet service and competition from terrestrial and other satellite service providers, so prices will vary within the U. S. and globally. In nations where Starlink service is sold by partner Internet service providers, they will share in pricing decisions.

Since the marginal cost of serving a customer is near zero as long as there is sufficient capacity, we can expect lower prices in a poor, sparsely-populated region than in an affluent, densely-populated region. Dynamic pricing is also a possibility since SpaceX will have real-time demand data for every location. "Dynamic pricing of a zero marginal cost, variable-demand service" sounds like a good thesis topic. It will be interesting to see their pricing policy.

National governments will also have a say on pricing and service. While the U. S. will allow SpaceX to serve customers directly, other nations may require that they sell through Internet service providers and some -- maybe Russia -- may ban Starlink service altogether.

The price and quality of service also impact long-run usage patterns and applications. Today, the majority of users in developing nations access the Internet using mobile phones, which limits the power and range of applications they can use. Affordable satellite broadband would lead to more computers in homes, schools, and businesses and reduce the cost of offering new Internet services, impacting the economy and culture and leading to more content and application creation, as opposed to content consumption.

Looking further into the future, SpaceX has FCC approval for around 12,000 satellites and they recently requested spectrum for an additional 30,000 from the International Telecommunication Union. Their next-generation reusable Starship will be capable of launching 400 satellites at a time and they will have to run a regular shuttle service to launch 42,000 satellites as well as replacements since the satellites are only expected to have a five-year lifespan. (One can imagine Starships dropping off new satellites then picking up obsolete satellites and returning them to Earth).

This sounds rosy. As we said in the NSFNet days, what could possibly go wrong? SpaceX seems to have a commanding lead over its would-be competitors. Might they one day become a dominant Internet service provider in a nation or region and abuse that position? Also, before they launch 42,000 satellites -- or even 12,000 -- SpaceX better come up with a foolproof plan for debris avoidance and mitigation. I hope they have a vice-president in charge of unanticipated side-effects.

Update 11/5/2019

Speaking at an investment conference, Shotwell said that a single Starship-Super Heavy launch should be able to place at least 400 Starlink satellites in orbit. Doing so would reduce the per-satellite cost to 20% of today's 60-satellite launches.


Update 11/6/2019

Serge Eagleson informed me that since Mark Handley ran his simulations, SpaceX modified the configuration of the 550 km shell in order to sooner serve the southern U. S. The new configuration will have 72 orbital planes of 22 satellites rather than 24 orbital planes of 66 satellites -- 1,584 sats either way. The change will broaden coverage in the southern U. S., but thin overall capacity. Serge ran a simulation of the tentative configuration by around the middle of next year, 18 planes of 20 satellites:

U. S. overage with 18 20-satellite planes

The following shows global coverage with 18 and 36 planes of satellites:

Global coverage with 18 and 36 20-satellite planes

At 18 planes, the Earth outside of the polar regions is nearly covered. With 36 planes, it is fully covered and there is more capacity (deeper blue shading). The remaining 36 planes of the shell will further increase capacity.

Update 1/16/2020

Gwynne Shotwell predicted a Starlink launch every two weeks, and they seem to be achieving that cadence. The next Starlink launch will be as soon as January 20 -- two weeks after the last one. That works out to about 1,500 launched by the end of this year. If they can maintain that rate, they will have completed their first 550-km shell around February 2021. (There will be some satellite failures, so that may take till March). At that point, they will have continuous coverage between about 53 degrees north and south latitude.

The financing also seems to be working out. Last May, Elon Musk said their recent fundraising rounds “have been oversubscribed” and “At this point, it looks like we have sufficient capital to get [Starship] to an operational level.” I don't know what an operational level means, but if it means 400 satellites per launch, we are going to be seeing a lot of Starlink satellites in the next few years.















Friday, September 06, 2019

Inter-satellite laser link update

SpaceX satellite mesh with four
laser terminals on each satellite
Inter-satellite laser links (ISLLs) and electronically steerable flat panel antennas are critical technologies for constellations of low-Earth orbit (LEO) Internet-service satellites. Low-cost antennas are critical for the mass consumer market and ISLLs are required for an effective Internet backbone in space. In an earlier post, we saw that progress is being made on antennas, this one looks at ISLLs.

The figure to the right is taken from a simulation of the first phase of SpaceX's planned broadband Internet service, Starlink. It shows 66 satellites in each of 24 53-degree orbital planes -- a total of 1,584 satellites at an altitude of 550 km. Each satellite has four laser-communication terminals. Two on the front and back and two on the sides. Since the front and back lasers link to satellites in the same orbital plane, they remain at the same place in the sky relative to each other while the side lasers must move to track one another. (To visualize the dynamic nature of the links between the constantly moving satellites, check this clip from the animated simulation).

When Elon Musk introduced his Starlink plan to prospective employees in 2015, he said his goal was to transport "a majority of long-distance Internet traffic" and "about 10 percent of local consumer and business traffic." He pointed out that satellites would have an advantage over terrestrial links since the speed of light is faster in space than through optical fiber and fewer router hops would be needed to reach a distant location.

In addition to mitigating the digital divide by serving rural areas and small organizations, Musk and his competitors at OneWeb, Telesat, Amazon, and Leosat hope to service high-end, high-margin customers like enterprises, governments and maritime, airline and mobile phone companies. ISLLs are necessary for serving those lucrative high-end markets.

Initially, SpaceX proposed five ISLLs for each satellite -- the fifth would have been a link to a satellite in the crossing plane, but last November they cut back to four. The fifth terminal would have been difficult to engineer because while the front, back and side-mounted terminals move slowly relative to each other, this simulation shows that satellites in crossing planes would have been traveling at 7.3 km/second relative to each other. Furthermore, links between satellites in crossing planes would be of short duration. Designing and manufacturing them would have taken time and money.

Furthermore, because of the 53-degree orbit inclination, about half the satellites are moving northeast and half are moving southeast at any time and place. That favors east-west links over north-south links and since most of the lucrative low-latency, long-link traffic is in the northern hemisphere, they could not justify the cost or possible deployment delay. That is not to say they will not deploy them in the future. (Note that the initial five-link constellation was to orbit at an altitude of 1,100, not 550 km. Future plans call for constellations at 1,100 and 335-345 km and there may be ISLLs between all of them).

Tesat laser communication roadmap
But even with 4-links, the terminals are still under development and will be expensive. At least two companies are working on ISLLs, Mynaric and Tesat.

Tesat already markets a laser communication terminal for LEO to ground transmission from CubeSats. Their CubeLCT is 9 x 9.5 x 3.5 cm, has a mass of 360 grams, consumes 8 Watts of power and communicates through the atmosphere to the Earth at 100 Mbps, with a 1 Mbps channel from the ground to LEO. They are developing an ISLL terminal based on that experience and, judging from the diagram shown here, they are pursuing laser communication between the ground, LEO and geostationary satellites.

Mynarc has announced that their ISLL terminal, the MLT-80, will be available in high-volume production this year and both companies are working on faster terminals. A while ago, I suggested that SpaceX would probably develop their own ISLL, but last March, Bulent Altan, a former SpaceX Vice President, joined Mynarc as co-CEO and a few days later Mynaric announced that they had raised $12.5 million from mystery constellation customer. Might the mystery company be SpaceX? Might it be Amazon, which entered the race late and has enough money to pay for terminals or even buy a stake in Mynaric or Tesat? We will know soon because test satellites equipped with Mynaric’s terminals should be launched in late-2019.

The following are selected characteristics of their forthcoming ISSLs:

Mynaric Tesat
Link distance 4,500 km 6,000 km
Data rate (full duplex) 10 Gbps 10 Gbps
Target mass <20 kg <15 kg
Power consumption <60 W 80 W
Sources Tesat, Myarnic

The SpaceX simulation shown above was for satellites with 4 ISLLs, but SpaceX launched their first 60 satellites without the ISLLs and, as far as I know, has not said if forthcoming satellites will have them or not. Arthur Sauzay, a French environment and space lawyer has pointed out that SpaceX argued for the allocation of radio frequencies for ISLs in a comment to a recent Whitehouse report on the impact of emerging technologies and their impact on non-federal spectrum demand, but they seem too large, heavy and slow to support a LEO network with long-distance, low-latency links.

OneWeb has decided not to use ISLs in their first constellation and will route traffic through terrestrial gateways. This decision seems to have been at least partially motivated by Russian insistence that satellite traffic passes through gateways within their borders. I imagine China and other nations will impose the same restriction.

Telesat remains committed to ISSLs, but say they will have the flexibility in their network-control system to route traffic coming to a country over satellite or terrestrial links. Erwin Hudson, vice president of Telesat LEO is confident that ISLs will be cheap enough to allow them to compete successfully with terrestrial fiber and 5G, offering fast, 30 ms latency broadband. They also have a $2.8 million contract to study inter-satellite laser links between their constellation and Blackjack, DARPA's 20 LEO satellite constellation and they are collaborating with Google on software, so we might see laser links between Telesat satellites and Google's balloons.

LeoSat is unique in that they are not pursuing the consumer and small organization markets, but are focused exclusively on large, high-end customers. They will provide fast, low latency, encrypted, reliable point-to-point connections to governments at up to 1.2 Gbps with latency under 50 ms and they have over $1 billion in pre-launch customer agreements. ISLLs are mandatory for the markets they are pursuing and since two geostationary satellite operators, Jsat and Hispasat, are investors in LeoSat, they may very well link to them in the future to offer a service similar to the SpaceDataHighway of Airbus and the European Space Agency.

China's Hongyun LEO broadband project is an ISLL unknown. China is doubtless working on laser communication in space, but I have no idea whether or not they will use it in their broadband constellation. Since they say the goal of the project is to serve rural China and they regulate Internet links to the outside world, Hongyun satellites may serve exclusively as "bent pipe" relays between rural locations and China's terrestrial network.

ISLLs will be needed if the Internet backbone in space is to compete with the terrestrial backbone and serve high-value applications. It seems that making cost-effective ISLLs for LEO constellations was harder than Elon Musk and others anticipated, but first production models are now on the horizon and they will improve over time.

For a copy of the PowerPoint presentation I use for teaching this topic click here.



Tuesday, August 06, 2019

An optimistic update from Telesat

Once the 100 inclined-orbit satellites are in orbit, they may be able to utilize their inter-satellite laser links to achieve the 30 ms latency Goldman spoke of.

Polar (green) and inclined (red) orbits
Emily Jackson interviewed Dan Goldberg, Telesat President and CEO, in a recent episode of the Down to Business podcast. The interview followed the announcement that the Canadian Government would contribute $85 million (all amounts are in Canadian dollars) to support research and development in support of Telesat's planned constellation of low-Earth orbit (LEO) satellites and another $600 million to subsidize Internet connectivity in rural Canada.

Goldberg pointed out that all governments subsidize rural connectivity and said the $600 million grant was expected to generate $600 million in revenue from below-market-rate sales to telephone companies and ISPs. The remaining capacity would be sold to others and he said they anticipated sales to enterprises, governments, ships, and airlines, but did not mention marketing directly to consumers. (Only SpaceX seems to be targeting consumers from the start).

In return for the R&D contribution, Telesat has agreed to support approximately 500 professional jobs in Canada and invest $215 million in R&D. (That R&D includes the first dozen or so test satellites). Telesat has a profitable, established geostationary satellite business and will fund part of the constellation themselves, but they will also need debt and equity financing and Goldberg said this government support would make it easier for them to finance the constellation.

This financial news is important, but Goldberg's optimism about the technology is what caught my attention. They have been working on their LEO project for six years and during that time the cost of launching satellites -- geostationary as well as LEO -- has fallen dramatically and he expects it to continue to do so. He also predicted that the cost of mass-produced satellites will fall dramatically and he is confident that inter-satellite laser links (ISLLs) and electronically-steerable phased-array antennas will be cheap enough to allow them to compete successfully with terrestrial fiber and 5G, offering fast, 30 ms latency broadband. (ISLLs present both technological and political problems).

The only technological concern he expressed was with regard to the problem of radio interference. He did not say anything specific on these technologies but did point out that Telesat has been providing satellite service for 50 years and is the "leading satellite technical consultant" in the world. (Three percent of their revenue is from consulting).

Goldberg summed up his optimism by saying:
Our confidence level in terms of our ability to bring this disruptive capability to the market and provide an extraordinarily high-quality, disruptive broadband service to Canadians and also to everybody else living in the world is extraordinarily high. This is not some high, big-gamble, futuristic new technology. This technology will be disruptive but it is ready for prime time.
Yes, but ...

SpaceX simulation with uncovered areas
Goldberg said they could could achieve global coverage with only 72 satellites and a simulation by Mark Handley predicts that SpaceX will not completely cover the planet with 792 satellites. How do we explain the difference?

SpaceX with 792 satellites would have much more capacity than Telesat with 72 satellites and Telesat does not plan to offer service with only 72 satellites. They plan to start service at the end of 2022 with around 200 satellites in polar orbit. They will add 100 more in inclined orbit in 2023 and perhaps eventually reach 500 satellites. Those 200 polar-orbit satellites will serve the polar regions, fulfilling their promise to provide connectivity in rural Canada. (This is reminiscent of China's Hongyun LEO satellite project which will focus on rural China).

While the 200 polar orbit satellites will provide coverage in rural Canada, they will be partially reliant upon terrestrial ground stations to reach the entire globe and therefore latency will suffer and they already have two far-north ground stations in support of their established, profitable geosynchronous satellite business. Furthermore, in 2016 Telesat filed for a patent on a "Dual LEO Satellite System and Method for Global Coverage" and once the 100 inclined-orbit satellites are in orbit, they may be able to utilize their inter-satellite laser links to achieve the 30 ms latency Goldman spoke of.

Friday, July 19, 2019

Latecomer Amazon will be a formidable satellite ISP competitor

Amazon CEO Jeff Bezos
In spite of being a latecomer to the race to deploy a constellation of low-Earth orbit (LEO) broadband Internet satellites, Amazon's Project Kuiper will be a formidable competitor. SpaceX, OneWeb and Telesat already have test satellites in orbit, but Amazon has several strategic advantages.

For a start, each of the LEO broadband competitors plans to end the digital divide by providing global connectivity to end-users and small organizations in underserved areas, but they are also counting on high-margin customers -- governments, enterprises, financial institutions, telephone companies, airlines, maritime companies and luxury yacht owners for early revenue. (A fifth company, LEOSAT, will focus exclusively on these commercial markets). Amazon's complementary infrastructure will give them a strategic advantage with these early customers. They will be able to leverage Amazon's established global Web and database services as well as their newly launched satellite ground-station service all of which will be integrated with the Project Kuiper constellation. Furthermore, when new end-users come online, they will be potential Amazon retail customers regardless of their satellite ISP.

The high-margin applications require inter-satellite laser links (ISLLs) for fast, secure long-distance communication and that technology is still under development. OneWeb has decided to forego ISLLs for their first constellation and SpaceX launched their first 60 satellites without them and, as far as I know, has not said when they will be deploying satellites with ISLLs. Amazon may be working on their own ISLL technology or planning to partner with (or buy) Mynaric or one of the partners in the European project ORIONAS (Lasercom-on-chip for next-generation, high-speed satellite constellation interconnectivity). Note that there are political as well as technical barriers to ISSL deployment.

SpaceX and OneWeb have talked of consumer ground stations costing as little as $200, but that will require another critical technology that is still under development -- cheap, mass-produced, electronically-steerable antennas the size of a "pizza box". Telesat says they will concentrate on the maritime, aviation and cellular-backhaul markets until the cost of end-user antennas comes down. SpaceX is developing their own antenna and has filed for permission to deploy a million end-user ground stations but an engineer working on the project told me they do not yet have an antenna that is cheap enough for the consumer market. OneWeb CEO Greg Wyler claims to have a self-funded side project that has developed a suitable fifteen dollar antenna and they may be ready to deploy. I don't know whether Amazon has been working on small electronically-steerable antennas internally, but even if they have not, as with ISSLs, they have the funds to either partner with or purchase a company that is working on them.

Debris mitigation is another technology for which no one has a proven lead over Amazon at this time.

Amazon also gained ground on the others when Elon Musk reportedly became frustrated with the pace of development at Starlink and fired the vice president in charge of the satellite program, Rajeev Badyal, a veteran of Microsoft and Hewlett Packard and satellite designer Mark Krebs, who led Google’s aircraft and spacecraft teams before coming to SpaceX and playing a key role in developing their first two test satellites. Amazon subsequently hired Baydal, Krebs and other ex-SpaceX engineers. I wonder if they influenced Bezos' decision to proceed with Project Kuiper.

Amazon has its own launch capability, but SpaceX has a clear lead in launch technology and capacity. Still, OneWeb has contracted with Amazon for five launches of perhaps 400 satellites starting in 2021 and one could imagine SpaceX serving their competitors as well. (I wonder if anti-trust law would require some sort of arm's length pricing).

Amazon CEO Jeff Bezos has deep pockets so will not have to worry about raising money and, perhaps more important, he will have complete control over the project. SpaceX has had to go to the capital markets several times, OneWeb is working with a group of investors and collaborator/investors and Telesat has income from its established geostationary satellite business, but is owned by a somewhat contentious combination of Loral Space and Communications and a Canadian pension fund.

Finally, Bezos has had the skill and vision to build an array of highly successful, complementary companies from online retail to fulfillment infrastructure to Internet services to space. That is not to take anything away from the others -- I suspect they were less surprised than I by the announcement of Project Kuiper. Whatever led to Amazon's decision, it is good to see them involved in a competitive battle among would-be global Internet service providers.

Update 7/22/2019

Megaconstellations points out that as a smart follower Amazon will also benefit from a matured ecosystem of suppliers and service providers facilitating mass production created and paid for by the first movers, OneWeb, Telesat and SpaceX.

Saturday, June 01, 2019

Amazon's AWS Ground Station service is now available

Amazon announced that they would be providing satellite ground station service last year and Andy Jassy, CEO of Amazon Web services, announced its availability in the video at the end of this post.

AWS Ground Station is a fully managed, ready-to go ground station service, featuring:

  • No upfront cost.
  • Scaleability -- you only pay for antenna time.
  • No long-term contract.
  • Self-service scheduling on a per-minute basis, that can be changed dynamically using their ground station console.
  • Secure transmission.
  • Low latency due to proximity to Amazon data centers.
  • Integration with EC2, S3 and other Amazon services and Amazon's global network backbone.
  • Simultaneous up/download.
  • Support of most common communication frequencies.
This sounds like a compelling case, especially for a small operator or startup, but I don't know how the prices compare to existing services or building proprietary ground stations.

A couple of questions come to mind. I assume Project Kuiper, Amazon's proposed broadband satellite venture, will use this service, but will SpaceX, OneWeb, Telesat and other potential satellite broadband ISPs also use it? If so, will Amazon treat them fairly? Competing ground station companies might also raise the issue of predatory pricing since Amazon will have an opportunity for cross-subsidy with their other services or they might just operate at a loss until competitors are eliminated (as they have done in other cases).

Hongyun Project -- China's low-earth orbit broadband Internet project

It might be tempting to dismiss this effort as small and behind the broadband satellite projects of companies like SpaceX, OneWeb and Telesat, but that would be a mistake.

Long March 11 rocket and Hongyun-1
satellite (source).
Last December, State-owned China Aerospace Science and Industry Corporation (CASIC) launched the first experimental Hongyun (rainbow cloud) Project satellite and they began testing it in March.

The 247 kg test satellite is in orbit at an altitude of around 1,100 km and they plan to launch four more test satellites this year and begin operating with a 156-satellite constellation in 2022. I don't know anything more about their plans, but with only 156 satellites I suspect they will focus on unserved regions in rural China and perhaps Latin America at first.

It might be tempting to dismiss this effort as small and behind the broadband satellite projects of companies like SpaceX, OneWeb and Telesat, but that would be a mistake. China has an ambitious, global Internet infrastructure and application program called the Digital Silk Road and the "road" is terrestrial with highways, ports, pipelines, and railways, undersea with cables and in space with the Hongyun Project, their Beidou satellite navigation system, which will be global next year, and the Digital Belt and Road Earth observation program. Our withdrawal from the Trans-Pacific Partnership and the current trade war were gifts to the Chinese.

(Other early short articles on the LEO project here and here).

Update 6/4/2019

CASIC broke ground on April 24 for a satellite industry park in Wuhan, Central China's Hubei Province, where they will produce satellites for the Hongyun project.

In keeping with China's policy of funding competitors, another production line operated by a satellite start-up, Spacety, based in Changsha, Central China's Hunan Province, began construction in January. Each facility is expected to produce 100 satellites per year. (China has historically funded Internet service competition).

Update 6/18/2019

U.S. military tracking data shows the satellite is in a nearly circular orbit averaging 1,067 km altitude at an inclination of 99.9 degrees and CASIC confirmed that Hongyun would emphasize service in China's remote regions.

Update 11/26/2019

Speaking at a conference last week, CASIC general manager Zou Guangbao confirmed their planned schedule and said they would serve the broadband communication, navigation & remote sensing markets in China and elsewhere. They are also developing a separate constellation of 80 Internet of things satellites

Update 12/14/2019

GalaxySpace is a second Chinese company working on a LEO broadband constellation. Their first satellite is under construction. The 200 kg satellite will have 10 Gbps capacity, orbit at 1,200 km with a 300,000 square km footprint and use high-frequency Q/V band radio.


Update 12/18/2019


Hongyun has expanded its broadband satellite plan. They are now working toward 864 satellites orbiting at 1,175 km with an 8 Terabytes per second capacity. They hope to serve 2 million 5G users through direct connections to base stations, 200,000 broadband users and 10 million Internet of things users. The focus will be on China and Belt and Road nations.

Update 12/21/2019

Liu Shiquan, Deputy General Manager of CASIC Hongyun satellite reported that Hongyun performance and function tests have been completed. He did not give details but said the tests included Web browsing, video chat, and high-resolution streaming.

Liu also gave a few schedule hints. The post quotes him as saying they would launch four more test satellites "by 2020" but I assume that meant "during 2020." He also said that by the beginning of 2020, users across China will be able to access the demonstration system. There was no elaboration on this, but I assume he is referring to a few test users. He also said they plan to have 156 satellites in operation by the middle of the 14th Five-Year Plan (2021-25).

I wonder if more detailed information is available on the Web in China.

Update 1/6/2020

The Hongyan (Wild Goose) project plans a constellation of around 320 LEO satellites. They have launched one test satellite so far and had hoped to launch 8 more by 2020, but did not make that deadline. They expect to have 60 satellites in orbit and operting around 2023 will be able to provide global coverage with the full constellation by 2025.

As shown in this illustration, they plan to connect buildings, ships, trains, and planes and to provide mobile backhaul and, most interestingly direct service to mobile phones. He Mu, Hongyan Application Director, promised the development of a "chip [that] can be integrated into the mobile phone so that everyone holding an ordinary mobile phone will have access to seamless satellite telecommunication with global coverage." (If this happens, it will be interesting to see how they differentiate this from terrestrial mobile service).

CASIC's Five Clouds
Hongyan is a project of China Aerospace Science and Technology Corporation (CASC) while Hongyun is a project of the China Aerospace Science and Industry Corporation (CASIC). CASIC has four other "five clouds" projects underway in addition to Hongyun:
  • Feiyun, using solar-powered drones
  • Kuaiyun, using near-space airships (dirigibles?)
  • Tengyun, a project to develop reusable space plane.
  • Xingyun, an 80-LEO narrowband IOT constellation using cubsats, the first of which has been launched.

Characteristics of Hongyun and Hongyan satellites














Update 1/20/2020

GalaxySpace has launched a LEO "5G" satellite, Yinhe-1, which is expected to test Q/V and Ka-band communications at up to 10 Gbps. I'm not sure what a "5G" satellite is, but note that the above diagram shows a satellite communicating directly with a mobile phone, as opposed to a mobile tower. Check out this short video on the satellite and launch:


Update 1/30/2020

See this post for some discussion of Chinese space policy.





Thursday, May 23, 2019

Might satellite constellations learn to avoid debris with sensors on satellites?

There were no artificial satellites before Sputnik in 1957.
Today there are about 5,000 with plans for thousands more.
Space debris is problematical.

The European Space Agency (ESA) reported that as of January 2019 there were about 5,000 satellites in space and 1,950 of them are still functioning. Hopefully, those functioning satellites have fuel and thrusters that will enable them to de-orbit and (mostly) burn up in the atmosphere when their useful life is finished. The remaining 3,050 are slowly drifting, along with a lot of debris.

The ESA estimates that there have been over 500 break-ups, explosions, collisions, or anomalous events resulting in fragmentation and they estimate that there are 34,000 debris objects >10 cm, 900,000 from 1 to 10 cm and 128 million from 1 mm to 1 cm. NASA says there are more than 20,000 pieces of debris larger than a softball, 500,000 the size of a marble or larger and many millions so small they can’t be tracked. (watch: NASA's Animation Shows Massive Space Junk Around Earth)

In low-earth orbit (LEO), debris circles the Earth at speeds of about 7 to 8 km/s. However, the average impact speed of orbital debris with another space object is approximately 10 km/s and can be up to about 15 km/s, which is more than 10 times the speed of a bullet. At those speeds, a collision with a small object can do significant damage. This sounds like a disaster waiting to happen and the current and planned proliferation of LEO satellites increases the likelihood of a Kessler Syndrome event -- a cascade of collisions between satellites and the ensuing debris.

As Kessler says "The cascade process can be more accurately thought of as continuous and as already started, where each collision or explosion in orbit slowly results in an increase in the frequency of future collisions." If you aren't worried yet, watch the following short video or read Kessler's 1978 paper.


Kessler's warning was taken seriously and NASA and others have been working on debris mitigation policy and technology for years, but the silver bullet has not been found. The Space Surveillance Network tracks approximately 23,000 relatively large objects and you can query the database here, but what about the millions of objects that are too small to track?

The SpaceX press release for their Starlink Mission hinted at their collision-avoidance strategy, saying that
Each spacecraft is equipped with a Startracker navigation system that allows SpaceX to point the satellites with precision. Importantly, Starlink satellites are capable of tracking on-orbit debris and autonomously avoiding a collision.
That sounds promising, but autonomously resolving and recognizing a marble-sized object that is approaching at up to 15 km/s, computing its trajectory and firing thrusters to avoid a collision can't be done -- even by Elon Musk.

Relatively few debris objects can be tracked terrestrially, but a satellite might be able to recognize a piece of debris and transmit its characteristics to a terrestrial processor, greatly expanding the tracking database. SpaceX may be approaching this as a machine-learning problem in which the entire constellation, not individual satellites, is learning to avoid collisions.

That is pure speculation, but it was triggered by a few thoughts.

For a start, at the end of 2017, SpaceX delivered a space debris sensor (SDS) to the International Space Station. As shown in the following short video, the SDS is capable of monitoring the size, speed, direction, and density of small particles that impact it.


Elon Musk also has a strong interest in machine learning -- he was a co-founder of openAI and his Tesla cars act as sensors uploading driving data that is used for training autonomous vehicles.

Going out to the very end of the limb -- Musk is a fan of science fiction and speculation on the possibility of a swarm of man-made objects learning about existential risks is reminiscent of emergent intelligence in Asimov's fictional planet Gaia or Teilhard de Chardin's noosphere).
Musk opened his Tesla patents and, if SpaceX demonstrates the feasibility of this approach to debris avoidance (and perhaps one-day removal), I expect that he would share this technology with competitors like OneWeb and Telesat and the space agencies of all nations.

Like global warming, space debris is an example of a tragedy of the commons and is a threat to all nations. As the cartoon character Pogo said, "We have met the enemy and he is us." Ironically, global tragedies of commons can unite us.

Friday, May 17, 2019

SpaceX reports significant broadband satellite progress

SpaceX may be approaching debris detection as a machine-learning problem in which the entire constellation, not individual satellites, is learning to avoid collisions.

Starlink size comparison -- novel packaging accomodates
60 satellites in a single launch. (Source)
SpaceX delayed last Wednesdays Starlink launch due to high winds and on Thursday they decided to do a software update and postpone the launch until next week, but they revealed significant progress in their Starlink mission press release and in tweets by and a media call with Elon Musk.

The mission press release said SpaceX has significantly reduced the size and weight of their satellites. Their initial November 2016 FCC filing specified 386 kg satellites that measured 4 x 1.8 x 1.2 meters. In February 2018, they launched two Internet-service test satellites -- TinTin A and B -- that measured only 1.1 x .7 x .7 meters with a total mass of approximately 400 kg. The mass of the Starlink satellites will be only 227 kg, about 43% that of the test satellites. (They are still heavier than OneWeb's 147.4 kg test satellites)

As far as I know, SpaceX has not previously commented on the number of satellites that might be launched at once, but the number was generally estimated as 25-30 after considering constraints on mass, volume, and numbers of satellites per orbital plane. As shown here, they will be launching a surprising 60 flat-packed satellites. Launching 60 satellites also demonstrates continued progress in rocket capability -- this will be the heaviest SpaceX payload ever.

The speed and density of satellites in
low-earth orbit increase the likelihood
of a cascading debris collision. Source
The current and planned proliferation of low-earth orbit satellites increases the likelihood of a Kessler Syndrome event -- a cascade of collisions between satellites and the ensuing debris. The satellites will be equipped with krypton powered thrusters that will enable them to autonomously avoid collisions with on-orbit debris that is large enough to track. The thrusters will also be used to de-orbit obsolete satellites causing 95 percent of all components to quickly burn in the atmosphere.

But, what about small, untracked objects? Low-earth orbit satellites move very fast and even if a satellite had the resolution and pattern-recognition capability to "see" debris in its path, it would not be able to maneuver quickly enough to avoid a collision. That point was raised in this online discussion and a possible solution suggested -- the entire constellation could dynamically pool and share data from each satellite as well as use NORAD tracking data, which Musk mentioned during the media call.

SpaceX may be approaching this as a machine-learning problem in which the entire constellation, not individual satellites, is learning to avoid collisions using its shared data as well as data from other sources like NORAD. One can imagine sharing such data with competitors like OneWeb and Telesat or even with Russia, China or India. (Elon Musk is known to read science fiction -- this speculation is reminiscent of Azimov's Gaia or Teilhard de Chardin's noosphere).

The prospect of launching 60 satellites at once and a shared-data approach to collision avoidance have grabbed my attention, but Musk's tweets and media call were also highly informative -- a few examples:
All that and they have yet to launch the satellites -- stay tuned.

Wednesday, May 08, 2019

Satellite Internet Service Progress by SpaceX and Telesat

This has been a busy week in the race to deploy constellations of low-earth orbit (LEO) Internet-service satellites.

Telesat LEO-1, artist's conception.
Credit SSTL.
In their quarterly report, Telesat mentioned progress in two, disparate markets. As I noted earlier, they have signed their first LEO customer -- Omniaccess a provider of connectivity to the superyacht market. Telesat is a Canadian firm and the quarterly report also said Canada's 2019 Federal budget included a commitment to using LEO satellite services to help bridge the digital divide. They will be serving Russian oligarchs and rural Canadians.

Telesat also announced that the two teams they contracted with to develop overall satellite and ground system proposals, Airbus and a consortium of Thales Alenia Space and Maxar Technologies, had significantly advanced their detailed designs for the complete system, having completed the system definition and risk management phase of the program. Telesat will continue their collaboration with both teams and will select a prime contractor later this year.

Telesat's coolest development was the announcement that they had demonstrated 5G mobile backhaul. They collaborated with Vodaphone and the University of Surrey in a test of their experimental satellites and recorded round trip latency of 18-40 milliseconds. The demonstration supported video chatting, Web browsing and simultaneous streaming of up to 8K video. The team also transferred 4K video to the edge of the 5G network. SES is already providing mobile backhaul using their middle-earth orbit satellites and it seems that the new LEO constellations will be competing with them. This will be an important application for rural areas and developing nations.

Telesat has signed launch contracts with Jeff Bezos’ Blue Origin and they plan to be operational by 2022. (Bezos will also be a Telesat competitor, but his LEO project has just been announced).

SpaceX also made the news. SpaceX president and chief operating officer, Gwynne Shotwell, confirmed that the launch scheduled for May 15 will include "dozens" of Starlink Internet-service satellites.

She characterized these as "demonstration" satellites and said they would not include satellite-to-satellite laser communication links. Bulent Altan, CEO of satellite laser company Mynaric, estimates that their laser terminals will cost around 250,000 euros in quantities of 1,000. SpaceX will have 4 in each satellite and they plan to start offering broadband service once they have 800 satellites in orbit -- in the 2020-2021 time frame.

OneWeb will forego inter-satellite optical links in their initial constellation, but they seem to be making steady progress in antennas for satellite-to-ground communication. Both are key technologies for LEO satellite Internet service.

Shotwell said that depending on how the demonstrations proceed, from two to six Starlink launches could follow by the end of this year. In the past, they referred to these as "operational" satellites. Maybe the switch to "demonstration" means they will use them as a marketing tool and the number of launches later in the year will be determined by sales.

Update 5/11/2019

TMF Associates suggests that the FCC may be favoring SpaceX Starlink and that the upcoming launch may carry as many as 40-50 satellites. The post also suggests that they may be launching so many satellites in order to generate publicity to spur further investment which has been difficult during the past year.

Friday, April 26, 2019

Google Plus was about community and collaboration and killing it damaged users

Anti-trust law keeps large companies from stifling competition. Could consumer-protection law keep companies from simply killing services that many people depend upon?

Google Plus chief architect Vic Cundotra "sold" the project to Google CEO Larry Page by convincing him that Facebook was an existential threat and, when Google Plus was launched, Facebook took it as an existential threat and responded accordingly.

Both were wrong, because, while they overlapped functionally, they were not direct competitors. Facebook serves a social feed of posts from family, friends, and people an algorithm identifies as being like you. Google Plus also offered a feed but was more about collaboration and support of communities of common interest.

J. C. R. Licklider, who was responsible for much of the research that led to interactive computing and the Internet, anticipated Google Plus Communities by over 40 years, writing that interactive communities would "consist of geographically separated members, sometimes grouped in small clusters and sometimes working individually. They will be communities not of common location, but of common interest."

I barely use Facebook but did use Google Plus. I didn't pay much attention to my Google Plus feed but used Google Plus Communities extensively. I have an interest in the Cuban Internet, so created a Google Plus Community on the topic and joined several other Cuba-related communities. I also joined communities on other topics I am interested in, like the Internet in developing nations and satellite Internet service.

I am also a teacher and created a Community each semester for my students to share material relevant to our class and study together. My students and I also used Google Plus Hangouts on Air for collaboration and coordination.

At the start, Google Plus was not just a social feed -- it was a collection of services that also included Photos, Hangouts, and Communities. Google separated Photos and Hangouts before they killed Google Plus and could have separated Communities as well, but they did not.

Would Google Communities have been viable as a stand-alone service? Yes. The closest competitor would have been Facebook Groups, but Groups lacks key features like post categories and the Communities user interface was far superior to that of Groups. Revenue sources could have included community-member data, ads, an optional membership fee, etc.

What about fake news, spam and toxic filter bubbles? It's easier to manage those things when the unit of scrutiny is a community rather than posts in a social feed. A community will have a creator and perhaps one or more moderators. They would be the first line of defense against inappropriate content within a community. The operator of the community platform would guard against communities that were intended to violate platform rules.

If they decided not to operate Communities as a separate service, Google could have offered the code and data assets for sale to others or at least put the code in the public domain. Evidentally it was easier for them just kill Communities along with the social feed, but doing so damaged me and other users. I don't know how many Communities there were or how many members they had, but killing Google Plus caused the loss of a large amount of social and monetary capital.

Over the years, Google has killed 137 services, 12 apps, and 12 hardware offerings, beginning with Google Deskbar after a three-year run in 2006. Anti-trust law keeps large companies from stifling competition. Could consumer-protection law keep companies from simply killing services that many people depend upon?

Such a law would not be all bad for Google. If constrained, they would still recoup some of the value of their investment and it would instill confidence in other companies that were thinking of offering products that depended upon them. If I were a developer, I would be reluctant to build a product that depended upon Google with its track record of killing 161 offerings in 13 years. Google Plus also had a symbiotic relationship with other Google services. For example, Google Plus drove some Blogger and YouTube traffic.

Elizabeth Warren and others have suggested breaking up large Internet companies. If we do so, let's not throw away all of the pieces.

Thursday, April 18, 2019

Open data leads to competition

When evaluating proposed mergers and breakups, control of data should be considered along with market impact.

In a previous post, I spoke of Amazon's use of customer and market data in restraint of trade, but they are not alone. For example, leaked internal documents show that plans to sell access to user data were discussed for years and received support from Facebook’s most senior executives. Facebook gave Amazon extended access to user data because Amazon was spending money on advertising and partnered with them on the launch of the Fire smartphone. In another case, Facebook discussed cutting off access to user data for a messaging app that had grown too popular and was viewed as a competitor.

As former FCC Chairman Tom Wheeler points out in a recent post, proprietary data is a source of market control and he cites two examples where opening data has led to competition. In the US, a law mandating open access to video content enabled satellite companies to compete with cable companies and in the UK, open access to customer banking data led about 200 organizations to offer new services in its first year.

Wheeler's position is elaborated in Unlocking Digital Competition, a report from the Digital Competition Expert Panel convened by the British Treasury Department. Their data-related recommended actions are:
  • Establishing data mobility and open standards between services: overcoming network effects which cause markets to tip by requiring systems to ‘talk’ to each other using open, standardised formats. This will mean consumers can port their data between networks, interact with users on other, similar networks, and smaller firms can plug their services into those of bigger ones. New business opportunities will open up that use, manage, and combine data made available. Consumers, in turn, will have new choices of digital services, with switching made much easier.
  • Securing access to non-personal and anonymised data: tackling the data barrier to entry for smaller and newer firms, while protecting privacy. The power of bulk data driving economies of scale and scope is a key reason new firms struggle to compete and bring innovative services to consumers. Overcoming this barrier will allow the digital economy to remain dynamic.
These are only two of the 20 recommended actions in the 140-page report. Those actions are grouped under six strategic recommendations for the government:
  • Sustain and promote effective competition in digital markets, by establishing a pro-competition digital markets unit, tasked with securing competition, innovation, and beneficial outcomes for consumers and businesses.
  • Take more frequent and firmer action to challenge mergers that could be detrimental to consumer welfare through reducing future levels of innovation and competition, supported by changes to legislation where necessary.
  • Update and effectively use tools against anti-competitive conduct to help them play their important role in protecting and promoting competition in the digital economy.
  • Continue to monitor how the use of machine learning algorithms and artificial intelligence evolves to ensure it does not lead to an anti-competitive activity or consumer detriment, in particular to vulnerable consumers.
  • Conduct a market study into the digital advertising market encompassing the entire value chain, using its investigatory powers to examine whether competition is working effectively and whether consumer harms are arising.
  • Engage internationally on the recommendations it chooses to adopt from this review, encouraging closer cross-border co-operation between competition authorities in sharing best practice and developing a common approach to issues across international digital markets.
The two open-data actions mentioned above fall under the first strategic recommendation of promoting competition, but control of data is involved in the others as well. When Amazon acquired Zappos and Whole Foods, they gained access to data on relatively affluent shoppers. Facebook's acquisition of Instagram and WhatsApp and Google's acquisition of Waze also yielded data in addition to eliminating competition. The machine learning recommendation involves training data. When evaluating proposed mergers and breakups, control of data should be considered along with market impact.

The last strategic recommendation -- international engagement -- recognizes the global nature of the Intenet. (Note that the Digital Competition Expert Panel was chaired by an American). Nations like China and the US have different goals with respect to competition, but democratic, capitalist nations should strive to adopt compatible institutions and policies. In the era of Brexit and MAGA, we need to work with other nations -- I'd rather end up with two Internets than fifty.

Thursday, April 11, 2019

Amazon's orbiting infrastructure

The invisible eye of the marketplace
I have been following satellite Internet service since the 1990s, but I was surprised when I learned last week that Amazon had filed an application for a 3,236-satellite constellation of low-earth orbit Internet service satellites -- Project Kuiper.

I shouldn't have been surprised -- Amazon was an infrastructure company from the start.

In his first post-IPO letter to shareholders in 1997, Jeff Bezos pointed out that their distribution center capacity grew from 50,000 to 285,000 square feet and said their goal remained "to continue to solidify and extend our brand and customer base. This requires sustained investment in systems and infrastructure to support outstanding customer convenience, selection, and service while we grow."

Today Amazon infrastructure is used internally and is offered as a service to others. Their distribution centers are now highly automated and they distribute a lot more than books. Amazon also offers Web, cloud storage, shipping and delivery services, credit cards, a voice application platform, an affiliate retailer program, satellite ground stations, automated retail stores, pickup locations, Whole Foods stores and other things I am probably overlooking. Bezos personally owns the Washington Post and the Blue Origin aerospace manufacturing and spaceflight services company. (Blue Origin has a contract to launch satellites for Telesat, a Project Kuiper competitor).

Bezos' preparation for Project Kuiper was hiding in plain sight with the reference to road building in the Blue Origen mission statement: "We're committed to building a road to space so our children can build the future" and it should have become more clear when Amazon added fully-managed satellite ground station service to its Web Service offering. Amazon says Project Kuiper "will provide low-latency, high-speed broadband connectivity to unserved and underserved communities around the world.” That may be true, but it is the tip of the iceberg -- like saying Amazon saying in 1994 that they would deliver low-cost books to homes.

During the industrial era, infrastructure companies like railroads and oil and steel companies grew as quickly as possible in order to achieve economies of scale and create barriers to entry and profit from usage fees and sales. In the information era, data is as important as fees and sales. Esther Dyson pointred that out in 1995, the year after Jeff Bezos founded Amazon, and it was reaffirmed recently when Softbank founder and CEO Masayoshi Son justified his billion dollar investment in OneWeb's satellite constellation, saying "whoever gets the most data wins."

Stacy Mitchel has researched the ways Amazon has applied Dyson's insight. Since many people go straight to Amazon rather than use a search engine when shopping for products, Amazon learns what people want, what they eventually buy and don't buy and how much they pay. They use that market knowledge to decide what to feature in search results, which products to brand or manufacture themselves, which companies to buy, etc. Their size and information facilitate optimal and, in some cases predatory, pricing. Mitchel cites book sales and their zapping of Zappos as examples of the latter and also shows ways in which Amazon has used government subsidy. (Imagine the price war between Jeff Bezos and Elon Musk).

Amazon harvests data from all of their infrastructure offerings. For example, Netflix uses AWS. They even learned a lot about local demographics, real estate prices, labor costs, etc. when they invited cities to apply to be the site of Amazon's future headquarters.

But, what's wrong with this? Amazon is efficient and has kept prices low and their customer service is terrific.

That sort of reasoning has dominated US anti-trust enforcement in recent years, but it is partial and short-sighted. Prices that are low enough to maintain rapid growth suit Amazon well now, but in the long run, competition and transparency fuel low prices, efficiency and the broad distribution of wealth and income.

Industrial era concentration of power resulted in anti-trust action in the early 20th Century, but those were simpler times. In the letter to shareholders mentioned above, Bezos also stated that "there are significant opportunities to better serve our customers overseas." How can we achieve competition between global Internet service providers like Amazon, SpaceX, OneWeb and Telesat that are outside the jurisdiction of a single nation?

There is no simple answer to that question. We can't put the genie back in the lamp, but we have seen some success with government operation of neutral, free or wholesale infrastructures like roads, sidewalks and municipal backbone networks and Google has had some success with fair, wholesale networking in Africa. Europe is beginning to look for ways to encourage competition, transparency, and privacy.

Update 4/12/2019

In yesterday's 2018 letter to shareholders, Jeff Bezos reported that 58% of Amazon.com retail sales are made by independent third-party sellers, emphasizing that they "helped independent sellers compete against our first-party business by investing in and offering them the very best selling tools we could imagine and build," i. e., infrastructure. He singles out Fulfillment by Amazon, the Prime membership program, Amazon Web Services, database tools, SageMaker for machine learning and Alexa, which is built into 150 different products.

He also points out that Amazon remains a small player -- "low single-digit percentage" -- in global retail largely because nearly 90% of retail remains offline, in brick and mortar stores. But they are working on infrastructure for them -- Amazon Go stores.

He has convinced me -- forget retail, Amazon is an infrastructure company.

Amazon's infrastructure yields revenue and data, and that data can be used in restraint of trade, but Amazon is not all bad. As Bezos points out, they create jobs and valuable services as do others who use their infrastructure. Elizabeth Warren and others are calling for changes to US tax and anti-trust laws, but we need to be careful not to throw out the baby with the bathwater. (Full disclosure -- I loved Mrs. Maisel, the Big Sick, and many more and bought a terrific pasta cooker at Amazon.com yesterday).

The issue is further complicated by the fact that this is not a US issue, it is global. Amazon (and others) are Global infrastructure companies and China is a global infrastructure country.

Update 1/13/2020

Amazon competes with and gathers data from retailers who host their stores with Amazon. A similar situation exists in satellite Internet service where both Telesat and OneWeb have signed launch contracts with Bezos' Blue Origen. Will SpaceX also launch satellites for competitors and what sort of scrutiny is necessary in these cases?