Starlink Fast Lane? | Hacker News Reader

Starlink Fast Lane?

3 pointsncmncm7 comments

SpaceX is, lately, lofting satellites with laser links to other satellites, I assume pointing them at others leading and following in the same orbit, so they don't need to track.

An orbit near 70° inclination takes satellites over New York and London. This enables Starlink to deliver traffic between those spots several milliseconds ahead of literally anybody else. Similar should be possible between NY or Europe and Asian markets.

Banks have spent many tens of $millions on microwave towers to get financial data between NYC and CHI just a few ms ahead of competition stuck on fiber 30% slower.

I wonder, is SpaceX offering these low-latency connections to banks for the big bucks? And, charging even more not to offer the same to competitors, too? Would the FCC allow the latter?

View on HN

In principle a satellite link like that could run at the speed of light in vacuum as opposed to the much slower speed of light in an optic fiber. You'd save enough that you could afford to detour 300 miles up to get to the network and 300 miles back down.

In practice it is tough to realize because the satellites are moving and different packets will follow different routes. Getting that last bit of performance consistency would be difficult if not impossible.

In fintech we say a microsecond is an aeon and a millisecond is an eternity.

If it gets there a millisecond before anybody else has it, that's all you need. And, anyway, there are only two routes: up 340-500 mi to this satellite to the next in front of it to the next and down; or back.

And, you don't need consistency: each message that arrives well before anybody else gets it is an independent advantage.

But is it possible to know if you are actually beating people if your connection is not consistent? For example, if half the time you are a millisecond faster than competitors but the other half you are 2 Ms slower, it seems like you would average out behind unless that is enough time to check the speed before making a trade.

Actually, yes. Market data packets are timestamped as they leave the exchange, so you know exactly how many microseconds they took to get to you. You also know exactly how long they would have taken to, say, cross the Atlantic via fiber. If a copy of the packet already got to you (and everybody else) via fiber, you also know that.

Finally, when your through-free-space data channel is reliably 10 ms faster than a fiber link, varying by a millisecond or five this way or that doesn't make much difference. Five ms of lead time is as good as 15. All you might not know immediately is whether somebody else is also getting packets early via Starlink. But it doesn't take long to find out.

Yes, but a trade has to go both ways. I can receive the information early, then send a trade to act on that information. But if my connection varies enough on sending to make me sometimes faster, but sometimes slower than the competition, I don't see how you could come out ahead.

You do local trades according to your better information, before others who don't have it yet. Those take only microseconds.

I see, that makes sense. I hadn't considered the information coming from a different source than where the trades happen. I was in the mode of thinking about trade front-running where you are trying to beat orders in progress.

Reply on news.ycombinator.com