Internet Satellite Links by SpaceX -- We Have a Few Questions

The Washington Post recently published an article: "SpaceX wants to beam the Internet down to Earth. Here's How it will Start."

IWL believes the article doesn't give the whole story.

SpaceX is not proposing anything that is particularly new.

Low Earth Orbit (LEO) satellite networks have been around since the 1990's. There was Iridium and the former Soviet Union had a LEO satellite network for voice calls to remote locations (like Siberia).

Low Earth Orbit (LEO) satellite networks are attractive because they avoid the long delays required for a packet to travel up to a geosynchronous satellite and then back down again. This delay can range from 240 to 280 milliseconds (source ground station to destination ground station.) On the internet, these are very long times.

When a satellite is in low earth orbit, the Internet delay from the ground is only about one millisecond each way. Consider a low earth orbit satellite that is 180 miles (289 kilometers) up from the surface of the earth. The propagation speed of light and radio is about 300,000 km/second. That works out to about a millisecond for a data packet to get up to a satellite, and the same to get back down.

SpaceX is looking higher - they are suggesting putting their satellites 1150km up. With a bit of trigonometry one can compute that their satellites will require roughly 5 milliseconds packet transit time, each way.

A big issue with LEO satellite networks is packet routing. Will SpaceX bounce the packets around from satellite to satellite? Or will the packets go to ground stations after each trip to a satellite? The latter is easy, but requires lots of hops up and down. The former is harder – not because of some complexity in the routing relationships between satellites - those are highly predictable and remain quite stable – but, rather, because satellites don't have a lot of spare electrical power. Radio power consumption increases with the square of the distance between the transmitter and the receiver. This means that the routing of packets between satellites tends to prefer to forward packets to nearby neighbor satellites rather than to distant ones. In other words, inter-satellite packet switching uses a lot of short hops. This is different from the traditional terrestrial routing in which the goal tends to be to use the fewest number of hops (and generally the longest links).

Satellites and ground stations often are reconciling multiple streams of traffic. And thus they have internal queues of packets that have been received and are waiting for their transmission time to arrive. Queues mean delay. And that delay usually varies depending on how long or short a queue may be when a a packet arrives. Variable delay is called "jitter".

As the number of hops increases so does overall end-to-end delay and jitter. Absent careful packet scheduling and expensive overprovisioning of satellites and radio resources we can anticipate that LEO packet switching networks may well have troublesome delay and jitter characteristics. LEO based packet switching networks may have to be vary careful to reject offered traffic that could cause congestion.

A common assumption is that satellite networks are clean and error-free. That is not always true. For example, there is the weather, specifically rain, which is the enemy of many satellite up/down links. And there are other causes of noise. Some are predictable, such as when a satellite or ground station is blinded by the reflection of the sun off of the earth or a satellite passes in front of the sun. Other causes are less predictable, such as bursts of solar radiation.

Traditionally the ground stations have been large, fixed location affairs. That will remain the case if our mobile devices attach to a hard-wired circuit, Wi-Fi link, or cellular signal. However, as our computer devices become increasingly mobile and increasingly connect directly to the LEO satellites, the ground stations - and especially their antennae - will have to become smaller and not require physical aiming. These are not insolvable problems; small, inexpensive, portable phased array antennae devices have been around since the early 1990's.

Because of these issues, this kind of LEO packet switching service is likely to become a kind of premium offload, i.e. something for traffic sources/sinks willing to pay premium fees. And because of delay and jitter issues, such networks may not be the best choice for delay+jitter sensitive conversational traffic, such as voice or conferencing. Inevitably, the question of costs and who makes the choices will become part of our ongoing "network neutrality" debates.

Finally, LEO satellites don't last long; they fall back into the atmosphere. Replacing them requires continuous pattern of launches. That is something that SpaceX will do very well.