HotNets'18: Networking in Space

HotNets'18 was held at Microsoft Research, Building 99. This is walking distance to my office at Cosmos DB, where I am working at my sabbatical. So I got tempted and crushed this workshop for a couple sessions. And oh my God, am I happy I did it. The space session was particularly very interesting, and definitely worth writing about.

My God, it is full of satellites!

According to a 2018 estimate, there are 4,900 satellites in orbit, of which about 1,900 operational, while the rest have lived out their useful lives and become space debris. Approximately 500 operational satellites are in low-Earth orbit, 50 are in medium-Earth orbit (at 20,000 km), and the rest are in geostationary orbit (at 36,000 km).

The low earth orbit LEO satellites are not stationary and fast moving around the earth at 1.5 hour per rotation. We are talking about the lowest ring in this picture, where International Space Station (ISS) resides.

Since LEO satellites are close to Earth, this makes their communication latency low. Furthermore, if we take into account that the speed of light in vacuum is 1.5 times faster than in fiber/glass, communicating over the LEO satellites becomes a viable alternative to communicating over fiber, especially for reducing latency in WAN deployments.

When this gets built, it will change Internet: in some accounts up to 50% traffic may take this route in the future.

And, it is actually getting built soon.

Starlink: SpaceX's satellite constellation

Starlink is a satellite constellation development project underway by SpaceX, to develop a low-cost, high-performance satellite bus and requisite customer ground transceivers to implement a new space-based Internet communication system. Development began in 2015, and prototype test-flight satellites were launched on 22 February 2018. Initial operation of the constellation could begin in 2020 with satellite deployment beginning mid 2019.

In Starlink’s initial phase, 1,600 satellites in 1,150 km altitude orbits will provide connectivity to all except far north and south regions of the world. A second phase adds another 2,825 satellites in orbits ranging from 1,100 km altitude to 1325 km, increasing density of coverage at lower latitudes and providing coverage at least as far as 70 degrees North.


And guess what! It looks like these satellites will communicate with each other using fricking ``lasers''!

Delay is Not an Option: Low Latency Routing in Space

Mark Handley (University College London) tried to reverse-engineer SpaceX's FCC filings to figure out what is possible with Starlink. It was the most interesting talk at the conference (at least among the talks I attended). People listened to the talk breathlessly and mesmerized. Mark had such exquisite visualizations and darkened the room for us to appreciate them better. It was a 20 minute trip to space and to 10 years in the future to deliberate about networking in space. (Here is a link to the paper.)

A special note about his slides is in order. He coded the satellites and the routing algorithms using the Unity framework. His slides were not showing video of simulations, but rather running the simulator in real-time. Bold and beautiful way to present.

The FCC filings mention "silicon carbide" communication components, which point to laser communication. Since it would be hard to infer bandwidth without more information, Mark took on the question of figuring out what the latency could be, and how would it change as satellites move, and what kind of use would this enable.

Each satellite has 5 inter-satellite communication links. The phase 1 satellites are northeast bound. And the phase 2 satellites are southeast bound.

The coverage is not uniform. London would be able to communicate with 30 LEO satellites at any given time.

Routing over satellites multihop via laser 90ms latency is achievable, compared to 160ms over fiber communication. This is a big improvement, for which financial markets would pay good money for.

Mark also considered how many multiple paths could be run over these satellites? He found that 10 multipaths is feasible. But 20 simultaneous paths not possible in phase1 of constellations.

With the additions in second phase (satellites that are southeast bound), London to Johannesberg latency can come down to 80ms from 190ms. These phase 2 additions will also help for providing better multipath capacity. With the second phase additions, FCC required SpaceX to cover the Alaska north region. This may also serve the purpose of routing over the pole, for example for the NY to Beijing route.

The Starlink deployments open many research questions for networking:

  • how do you avoid building queues? (probably via a form of source routing)
  • how do you coordinate multipath route changes?
  • how do you avoid reordering without increasing latency?
  • how do you make topology adaptive?

The other papers in the session

The "Networking, in Heaven as on Earth" paper considers the interdomain routing problem with satellite constellations. The vision there is to full integration of satellite networks in Internet Control Plane (via BGP). But satellites move very fast which leads to frequent BGP updates. Filtering reduces updates but introduces connectivity problems. The paper mentions that a proactive routing strategy (that leverages predictability of satellite orbits) rather than reactive could work better

The "Gearing up for the 21st century space race" paper talks about miscellaneous issues in space networking. The talk mentioned that some trade activity shows outstripping of fiber speed communications from NY to Frankfurt and that people might be using short-wave (microwave) radios to beat fiber-optical speeds (where light travels 2/3rds slower than it does in vacuum). Then the talk speculated whether it is possible to establish multihop microwave routing using in-flight planes. It turns out it is possible to do it with 20 hops across the globe (east to west) with low-stretch and good latency.

MAD questions

1. Maybe we are getting there, huh?
This session reminded me of the Seveneves book by Neal Stephenson (great read, recommend highly). In Part 1, of the book there was very good coverage of Space orbits, maneuvering in space, and how dangerous space junk could get. Coincidentally, one of the talks mentioned that space is garbage tracked: anything bigger than a marble is tracked. At first I didn't buy this, didn't sound feasible. But turns out radar is used for learning trajectories of the space junk and the trajectories are maintained at the databases in space agencies to help make the  space station and the satellites to avoid them. So the satellites will be routing packets and simultaneously try to route around occasional space junk. We are getting ready to become a space-faring species, and that is very exciting.

2. What is next?
Faster than light quantum communication, anyone? Ender's Game series mentioned such communication. And of course there is a wikipedia page for faster than light communication.

3. Would it be feasible to do store and forward communication via the satellites?
You know the thought-experiment about the plane full-of-disks, right? It has very good throughput. Since these satellites are already moving at a fast speed, could they be used for data mules to improve throughput for big data networking, say between the Hadron collider in CERN and datacenters in US? Think of beaming up data to a row of satellites (one after another) that store this data and in 45 minutes or so dump these at the US datacenters.  Could this be a feasible alternative to fiber? Probably not so much, since the uplink and downlink are still limited.


Anonymous said…
typo: space-fairing -> space-faring

the geometry of Starlink proposes interesting problems. See
Warren said…
Reminds me of something I wrote 2 years ago:

Combining internet routing with localized (and distributed) storage could be huge
Anonymous said…
Everything old is new again.

Popular posts from this blog

Graviton2 and Graviton3

Foundational distributed systems papers

Learning a technical subject

Learning about distributed systems: where to start?

Strict-serializability, but at what cost, for what purpose?

CockroachDB: The Resilient Geo-Distributed SQL Database

Amazon Aurora: Design Considerations + On Avoiding Distributed Consensus for I/Os, Commits, and Membership Changes

Warp: Lightweight Multi-Key Transactions for Key-Value Stores

Anna: A Key-Value Store For Any Scale

Your attitude determines your success