Hacker News Reader

Comment by tpmx | Hacker News Reader

I saw something interesting the other day that I had previously missed:

Starlink has the potential to provide lower latency around the globe than terrestrial fiber because light travels faster in vacuum than in optical silica fiber.

"Delay is Not an Option: Low Latency Routing in Space"

https://discovery.ucl.ac.uk/id/eprint/10062262/

http://nrg.cs.ucl.ac.uk/mjh/starlink-draft.pdf

"We conclude that a network built in this manner can provide lower latency communications than any possible terrestrial optical fiber network for communications over distances greater than about 3000 km."

View on HN

I wonder how the Starlink latency compares to the current microwave technologies used by high frequency trading firms. Any slight latency improvement is a game-changer in the HFT industry.

IIRC this is how they plan to finance it.

You just made that up.

I doubt that, since I've also heard this claim several times.

If he was lying why would he say IIRC?

Maybe he just doesn't recall correctly.

Why yes I make random shit up about how people plan to finance their low latency satellite constellations allla damn time. Ya got me. It's a fair cop guv.

I don’t recall SpaceX explicitly stating who they expect their customers to be, but fans quickly dredged up currency and remote stock trading as one potential major source of funding (single customers with huge budgets, rather than hundreds of homes on the cheapest plan)

This is a function of the permittivity of the material. A Vacuum is 1, the lowest possible. Air is 1.0006 so you won't see much difference.

Fiber optics of various types are usually on the order of 2 to 3 (closer to 2).

Distance is still a huge problem. The Starlink signal still needs to travel up to the satellites 500 KM up, and then another 500 KM back down. That should be around 2 ms (thanks user jobseeker990, corrected from previous 15ms) of added latency, so anything that is already below 4 ms of latency probably wouldn't be improved.

Would be fine with equivalency - that's a democratizing force. Will also spur more innovation in this area on behalf of HFT firms.

Light travels 1000km in under 4ms. What else are you adding to that time?

That's true in vacuum. Light traveling through a medium has to reflect off of or absorb/re-emit from everything it bumps into, so it goes slower.

I have updated my comment above. I was off by about an order of magnitude.

I think your math is way off. I'm getting less than 2 ms to go 500 km.

You're right, thanks. I was off by an order of magnitude. I've updated my comment

Lower latency across the globe is likely irrelevant to people that colocate in the same data centers as the exchanges to reduce latency.

Then again, maybe it's useful for arbitrage between different exchanges, where they would likely have equipment next to each exchange, and have a wide distance between them.

In case you didn’t know: there are several dedicated microwave links between Chicago and New York just for that reason.

I expect some early customers to be trading houses that want a faster connection for HFT.

Pretty neat. I guess they could vendor out to Google. GCP Spanner relies on reliable low ping fiber between zones so this might be a clear throughput win for them.

A high bandwidth dynamically directed laser-based mesh network of 12k-42k tiny satellites 340 km above the earth is a pretty ballsy concept to begin with.

I'm curious about who came up with this idea. It's kind of brilliant.

Arthur C. Clarke, in a way.

http://lakdiva.org/clarke/1945ww/

Sure, we owe Clarke a great deal of gratitude for that (the concept of the geostationary satellite 35k km from earth, from way back in 1945!), but this is something else and new.

It follows pretty logically from any kind of work in the wireless business. There’s a very direct correlation between transmitter altitude, and how likely it is that the transmitter will be able to connect to any given client. It follows that the best transmitter is one that is in space, which is why communications were one of the first commercial uses of satellites.

That's a way too simplified view that doesn't take latency into consideration. Or even the concept of a network.

I thought I read somewhere that they cancelled the plans to use lasers and were going to use radio instead. But now I can't find anything about it.

Radio and light have (for the purposes of this comparison) the same speed in vacuum.

They don't have FCC licensing for the satellites to communicate with each other using radio, the radio links are sat<->ground only.

FCC licensing in space...

The current generation of Starlink satellites don't have satellite-to-satilite links, but I haven't seen any indication that that feature is canceled for future iterations.

How do they work without sat to sat links? Are they going to build thousands of ground stations?

They have not cancelled the plans to use lasers, but have delayed the introduction of lasers until "satellite version 2.0", whatever that means.

That is, the initial constellation (my guess, the first 800 satellites) goes up without intersat links and acts as a bent pipe only. Lasers are still very much a plan for the later parts on the constellation, and the laserless satellites will eventually be replaced with ones with the lasers once they age out, but the laser links were simply not ready yet and their spectrum license essentially forces them to put the birds up on as quick a schedule as they can.

The laserless sats will not use radio to communicate with each other, but rather will use it to communicate with customers and fixed earth stations.

minor correction, they have delayed the laser interlink for version 1.0, which should be the next launch. the first launch is considered by spacex to be version 0.9 satellites

Yeah pretty interesting. The primary use-case I imagine that it will improve is voice and video real-time communications. Anything that lowers the round-trip time during a human conversation is going to be very significant to improving the user experience.

I guess high frequency traders might be interested too then.

Aren't they all just across the street form wall street with exactly the same network ping as everyone else?

That doesn't help if you're transmitting information from Shanghai Exchanges to London trying to gain first mover advantage.

Why would you do that?

Because there are billions of dollars/year of money in it.

Things that happen outside New York have an impact on stock prices in New York. Much of HFT is essentially companies racing each other to be the very first to act on external news/stock movements.

arbitrage

Because when something happens in China companies in the UK and the US that have supply chains there (almost everybody) will be affected.

The trades originate from elsewhere around the world.

Real Engineering has an interesting video on why it would be useful for firms trading internationally

https://youtu.be/giQ8xEWjnBs

Reply on news.ycombinator.com