Saturday, February 23, 2019

Rethinking technology policy - the 50th anniversaries of the moon landing and first ARPANET message

The anniversaries of these milestones remind us of the economic and social returns we have seen from ambitious Federal research, development and procurement programs. Why isn't this a Sputnik moment for the United States?

Log entry for the first ARPANET
message (Source)
On July 20, 1969, Neil Armstrong stepped out of the Apollo 11 lunar lander and set foot on the moon. On October 29th, Charley Kline, an ARPANET programmer at UCLA, attempted to log in to a host at Stanford Research Institute (SRI), but the system crashed after he had typed LO. After a bit of debugging, he succeeded in typing LOGIN and the rest is history.

These events were milestones in ongoing, federally-funded research and development efforts -- the manned spaceflight program and the development of digital telecommunication networks from Morse's telegraph to early electronic digital computers and telemetry experiments, the SAGE early-warning system and eventually the ARPANET and the Internet. (The previous link is to a text version of the history article. Contact me if you would like a pdf of the article as it appeared in the Communications of the ACM).

The anniversaries of these milestones remind us of the economic and social returns we have seen from Federal research, development and procurement programs. When President Franklin Roosevelt was inaugurated in 1933, he enacted many ambitious government programs aimed at ending the great depression -- a "new deal." He initiated programs like Social Security and infrastructure projects like rural electrification. Confidence in the Federal Government being able to pull off large, successful programs continued through World War II, the Berlin Airlift and Marshall Plan, the Interstate highway system and the space program.

Early ARPANET sketch, showing
UCLA and SRI nodes (Source)
President Kennedy kicked off Apollo program in 1961, saying he believed "that this nation should commit itself to achieving the goal, before this decade is out, of landing a man on the moon and returning him safely to Earth." But pendulums swing and, in a 1986 press conference, we heard President Reagan say “The nine most terrifying words in the English language are: I’m from the government and I’m here to help you.” Today we are discussing a "green new deal" -- is the pendulum about to swing back?

Update 3/14/2019

Former National Security Advisor Tom Donilon addressed technology policy in an terrific interview. The following is a paraphrase of his reflection on the emergence of collaboration between government, universities and private companies that was born when the Russians launched their Sputnik satellite:
I don't think we're having anywhere near the conversation we need to have about the missing piece of China policy which is: what is the United States going to do? Why isn't this a Sputnik moment for the United States? In the late 1950s, when the Soviets launched the Sputnik satellite, the United States undertook an enormous national effort that changed the way we taught math and established NASA and the Golden Triangle of technological and innovation development between the government and research universities and private companies.

We face a similar challenge at this point in terms of technology competition with China. That is something I think we need to be talking about and the discussion includes immigration policy, infrastructure investment, education investment and [consideration of the] impact of technology, particularly artificial intelligence, robotics and automation on labor markets. Populism is not at its peak right now -- we need a really serious discussion about how we are going to manage the future of work in the face of technology.
This comment was made at around 33:15 of the interview:

Friday, February 15, 2019

Google balloons and Telesat satellites

Telesat will use Google's network operating system. Will Google get access to data? A global backbone?

Loon balloons float at an altitude of around 20 km -- above
birds and the weather. They navigate by moving up or
down to catch wind currents moving in different directions. 
Telesat is making progress. Within the last month, they announced a launch contract with Blue Origin, a successful antenna test with Ball Aerospace and completion of system requirements reviews, but perhaps more interesting is an agreement to use the software defined network (SDN) platform Google has developed for their Project Loon.

Google has been interested in connectivity for underserved areas for many years, and have investigated and invested in non-terrestrial solutions using satellites, blimps, drones and balloons. Project Loon uses constellations of balloons in the stratosphere. They began working on it in 2011, provided emergency connectivity in Peru in 2016 and Puerto Rico in 2017 and are slated to begin commercial deployment in Kenya later this year.

Like LEO satellites, Loon balloons are in constant motion relative to the Earth and each other and the Loon SDN "schedules, predicts, controls and optimizes the wireless topology, radio resources, and routing of packets across the ground and aerospace segments of non-geostationary networks." The satellites and balloons are at different altitudes and move at different speeds, but the network characteristics are similar and Google will adapt their SDN for Telesat's satellite constellation.

Nothing was said about the terms of the deal. Since Google is also an investor in SpaceX, they now have an interest in two of the three large LEO satellite Internet-service projects. In addition to being paid to develop and maintain Telesat's SDN and getting a return on their SpaceX investment, they might be getting access to the data flowing through the networks or to a global backbone.

Softbank invested a billion dollars in OneWeb's satellite-Internet project for access to data. Softbank founder and CEO Masayoshi Son outlined his vision of the future in the keynote session of the 2017 SoftBank World conference. He believes the information revolution will be driven by strong, general artificial intelligence (AI), therefore the key material asset for the information age will be AI training data. His conclusion is "whoever gets the most data wins." Google uses data for AI training and for advertising.

Another intriguing possibility is that Google and Telesat might be planning to integrate their balloon and satellite networks. One can imagine Project Loon using Telesat's satellite network as a global backbone. That integration would be facilitated by their both running the same SDN software -- the same network operating system.

Whatever the motivation, this partnership provides Telesat with a strategic software asset and Google may gain access to data and a fast, global backbone.

For progress reports on the three LEO broadband projects, see OneWeb, Telesat and SpaceX.

Check this short Project Loon video:

Update 6/5/2020

9to5 Google found an interesting Google Careers listing for a “Partner Manager” whose role to “help launch a global satellite-based broadband service.” The partner will "support satellite broadband service providers, productize the solution and make it available to other satellite broadband ISPs.”

The ad does not specify which broadband ISPs the applicant is expected to support, but it could be SpaceX, where Google is an investor, Telesat, a network operating system customer, or both or others. One way or the other Google will win if LEO satellite broadband is a success. Amazon will win too.

Saturday, February 09, 2019

Fifteen-dollar, electronically-steerable antennas for satellite and terrestrial connectivity

Wafer antenna, credit Greg Wyler
What will the cost and capability of these antennas be in ten or twenty years? What would be the applications and implications of ubiquitous wireless networks? You might have some short-term guesses, like smart cities, but farther out it's science fiction -- Gaia, the Noosphere, 1984?

Three companies, OneWeb, SpaceX and Telesat, plan to provide global, low-latency broadband Internet service using constellations of low-Earth orbit (LEO) satellites. They anticipate millions of $2-300 end-user ground stations and that will require a critical technology -- cheap, electronically-steered antennas that can seamlessly switch from one satellite to another as they move across the sky. Antenna development has seemed to be a sticking point, but each of these companies has made a promising announcement during the last month.

OneWeb founder Greg Wyler announced that his self-funded side project, Wafer LLC, has developed a flat, low-power phased-array antenna that could be mass produced for $15. Wyler said tests of prototypes showed 50 Mbps capacity per antenna "tile" and said multiple tiles could be combined. (I assume that combing them would cut into the number of satellites that could be visible at any one time). The prototype uses the Ku frequency band that OneWeb's first satellites will use, but Wyler said they can be redesigned for different frequencies and orbit altitudes at little change in cost.

Telesat has two test satellites in orbit and Ball Aerospace announced a successful demonstration of communication between the test satellites and Telesat's ground station in Canada. They reported quick and seamless switching between the satellites and the speed and latency needed to stream video. (I asked about the characteristics of the video, but got no comment). Ball also says they will be able to mass produce their antennas. It is noteworthy that Telesat has said they would concentrate on the maritime, aviation and cellular backhaul markets until the cost of end-user antennas came down. It seems to have done just that.

SpaceX also has test satellites in orbit, and they report high bandwidth and 25 ms latency, but, unlike OneWeb and Telesat which are working with a number of suppliers and partners, they are vertically integrated, and are designing their own antennas. Earlier this month, SpaceX filed an application with the FCC for permission to deploy one million end-user Earth stations. The application describes their antennas as employing "advanced phased-array beam-forming and digital processing technologies to make highly efficient use of Ku-band spectrum resources by supporting highly directive, steered antenna beams that track the system’s low-Earth orbit satellites." In addition to steering the beams, they are able to vary power as a function of the angle between the ground station and satellite.

These antennas will enable Internet connectivity in homes, schools, clinics and other fixed locations. They will also be found on things that move like ships and planes. If all goes well, they will be on the roofs of cars and trucks as well (Teslas first).

They will play an even larger role in terrestrial communication. Since they have no moving parts and do not require precise alignment, they will be reliable and easy to install. They will become staples of 5G mobile installation, community networks, and other do-it-yourself projects. They will be deployed in dense urban areas, rural areas, developing nations, etc. (Years ago, one of my classes managed the networking of several of the dorms on our campus -- these antennas would have made that a snap).

What will the cost and capability of these antennas be in ten or twenty years? What would be the applications and implications of ubiquitous wireless networks? You might have some short-term guesses like smart cities, but farther out it's science fiction -- Gaia, the Noosphere, 1984?

For progress reports on the three LEO broadband projects, see OneWeb, Telesat and SpaceX.

Update 1/13/2020

User terminals as relays (source)
Mark Handley has simulated the first 1,584-satellite phase of SpaceX's Starlink constellation, assuming that there are no inter-satellite laser links (ISLLs) and all traffic is routed via terrestrial links. He found that even without ISLLs, latency was generally still better than in today's terrestrial Internet. This raises the possibility of designated user terminals being used as relay points when they are idle. Doing so would require user terminals with two (or more) antennas, but that might be feasible given the progress noted above. It would be interesting to scope out the cost and power requirements of a two-antenna user terminal. SpaceX could deploy thousands of such terminals, perhaps subsidizing their owners.