Is The Internet In Space An Idea That’s Really Down To Earth?

Blog By Dave Robinson Senior Network Architect (End to End Networks)

Elon Musk has recently announced building a network of satellites that will provide space Internet for the masses. The focus will be on providing Internet to the rural masses who can’t get it reliably or quickly today. This solves a real problem. My sister-in-law is a rural Canadian who can get satellite Internet, but the latency is massive and the bandwidth caps are barely better than mobile Internet. Now consider this: she’s one of the lucky rural Internet users. Many have no access at all.

Musk proposes that SpaceX will solve this problem. He’s also suggesting that he’ll build out an interplanetary Internet reaching to Mars. Keeping our feet on the ground though, Musk made an interesting statement about his satellite Internet. Aside from rural philanthropy, he’s also claiming that he’ll be able to achieve a faster Internet as light travels 40% faster in the vacuum of space than it does through fiber optic cables. That needs some unpacking.

While the speed of light statement is obvious, two other questions need consideration. How long does it take to the signal into space in the first place? Furthermore, granted that the latency will be 40% less in a vacuum, what is the wireless technology that Musk will use to achieve faster wireless throughput than transatlantic fiber optic cables offer today?

The first question is easier to answer. Satellite Internet providers today use satellites in geosynchronous orbit. This makes good sense because you simply point your antenna to the right place in the sky and as long you’ve got line-of-sight, you’ve got Internet access. Geosynchronous orbits sit some 32,000km above the earth though. Light takes about 120ms to get to those satellites, and another 120ms to get back. That makes for a whopping 240ms up and down — almost a quarter second. Many rural customers experience only that because they only care about download speeds. They have some low bandwidth land connection for upstream access. So in a typical use case, a tiny request for a web page goes through a land line that doesn’t need much bandwidth. It hits the ISP on the ground, the web page is retrieved, sent to the satellite (120ms) for transmission, and the satellite sends the web page down to the client who requested it (another 120ms). Some (mostly businesses) have expensive antennas on their roofs capable of sending data into the sky as well. Now that web page request takes 120ms to hit the satellite, another 120ms to come back to the ground where the web site is, another 120ms for the page to go back up to space, and a final 120ms for the page to be delivered from space to the customer. That’s 480ms of latency assuming that the satellites themselves and the rest of the Internet introduce no latency at all!

At 240ms round trip, the Internet feels terribly slow, no matter how much bandwidth one can offer. At 480ms, it’s more than crippling. The current satellite providers “fix” this by utilizing TCP accelerators that pull off some great trickery. The result is that while things seem slow to start with, once a session gets going, things seem pretty snappy and you can enjoy the bandwidth that’s being offered. In recent years, there have been some advancements in TCP itself to handle high-latency connections much better, so the accelerators may play a decreasing role in the future.

Still though, that’s only TCP. For everyone who builds a VPN connection to their office using UDP (NAT Traversal) or ESP (IPsec), you’re out of luck. Your traffic will seem dreadfully slow – if it even works at all. That goes for users using some of the newer P2P protocols out there too. And this is only going to get worse. Streaming video is going to use UDP rather than TCP more and more. Gaming is in the UDP boat too. Simply getting TCP to work just isn’t enough.

SpaceX has a great answer here. Rather than geosynchronous orbit, they are planning a network of satellites in low Earth orbit. Rather than 32,000km above the Earth, SpaceX will be a mere 1200km above the Earth. The one-way latency for that distance is only 4ms. For transmitting to space, then back down to the Internet, then receiving data back from the Internet and sending it to space, and finally from space back to your home (which clearly must be the goal here) we’re only looking at 16ms. Outstanding!

Now that the signal can get into space quickly it can take advantage of the 40% increase in speed to be achieved by sending data through space rather than fiber optic cables. It’s going to need it. The circumference of the Earth is about 40,000km. Once you’re 750km above the Earth though, the circumference of your satellite ring is almost 44,000km – a 10% difference in distance. That’s eating into the 40% speed increase.

The real issue is that our current transatlantic cables are capable of providing something on the order of 70ms to cross the pond. 40% of that is 28ms. If we could shave 28ms off of Internet latency, it would make great sense. But it’s really more like shaving off 30% because of the larger circumference, which makes our savings only 21ms. This is one-way end-to-end latency, which means in space that’ll include getting the signal to space and back down again, an 8ms hit. Now that 21ms becomes 13ms. Still better, but getting slim, and it sure ain’t 40%.

Finally, Musk needs a way for these satellites to transmit tens of gigabits per second wirelessly to rival current cable technology. He probably needs to be able to do it all with the power of the sun too, since these things will run on solar power. This is no easy task.

Complicating matters is the nature of low Earth orbits. Musk will need a network of satellites that will have to be talking to each other. Maybe this is easier in space, but this is tricky from the ground. Since the orbits aren’t geosynchronous, one cannot simply point an antenna to the right spot in the sky. There is no right spot. Relative to the Earth, the satellites keep changing position. The Earth bound antennae will need to track the satellites in space in order to keep in constant communication with them.

SpaceX thinks it’ll cost them about $10B to launch the entire project. That makes sense, since aside from the wireless space protocol that will send data at ten of gigabits per second, all of the other technology has already been made.  So why hasn’t this been done already. It has. Sort of.

Teledesic tried to do this 90s. Low Earth orbit, a network of satellites, high speed connections on Earth. It’s all been tried. The cost? Teledesic projected $9B. That’s $9B of 90s dollars. Irridium, Orbcomm, and Globalstar are all companies that tried something similar with low Earth orbits – though at lower bandwidth.

The Teledesic project is most interesting because it was partially funded by Bill Gates himself. So who’s funding this new SpaceX project? It’s partially funded by new tech giant on the block. Google and Fidelity are putting up $1B in exchange for roughly 10% of SpaceX.

It’s easy to look at this and imagine that it will be another failure. However, there’s something about betting against Musk that doesn’t sit well. If there’s someone who could actually make this happen, it’s the guy who’s bringing electric cars and reusable rockets into the mainstream. If they’ve managed to get $1B from Google and Fidelity, they might just be onto something.

This could mean that truly redundant Internet connections (one wired and one in space) will become common place. Just like we’re seeing the 3G Internet-on-a-stick making small waves today, we could soon see space transmitters and receivers making much larger waves tomorrow. This would be a new horizon for reliable Internet, as well as for network installers, troubleshooters, and managers.