Hacker News Reader

Comment by futevolei | Hacker News Reader

How are they getting latency so low? I used satellite internet for a month back in 2009 and it was brutal. Minimum 500ms. Literally had to break my lease to move to another place that offered broadband so I could actually work. I see starlink satellites are orbiting much lower but that would seem to only account for maybe 25% (?) of the gap between 20ms and 500ms.

View on HN

In this case its mostly the altitude. Old satellite internet talked to geostationary sats in orbit (out in space around 32000km), these are only around 500km, so the signal itself only has 1.5% the distance to travel. Just speed of light alone means your old signal took around 100ms to get to the sat, and 100ms back, new one is 1.5ms there and 1.5ms back.

Wow, I didn’t realize how much lower these satellites are.

It is also worth noting that propagation velocity in glass fiber is about 60% slower than free space. Once the constellation is dense enough and OISLs are working, they'll be able to undercut fiber latency. That is, if they actually decide to route signals in a purely latency-optimized manner. I wouldn't be at all surprised to see some sort of tiered pricing related to latency and latency jitter.

No doubt the scummy high frequency traders are already jockeying to get out in front of this change in the telecom industry.

I think your number is wrong right? Fiber should be 60% OF the speed of light, not 60% slower, so like 180km/s instead of 300km/s.

since we're being pedantic about fractions, I should point out that your figure for c is off by about three orders of magnitude.

What would be the use case for hft? I assume this would be slower than any existing microwave networks, or whatever is fastest right now. Cities that for whatever reason can’t be connected via microwave like nyc-London? I’m aware this possibly/probably reads like amateur hour to anyone actually in the hft biz...

Yes, routes like NYC-London or London-Johannesburg. http://web.archive.org/web/20200819084001/http://nrg.cs.ucl....

Show me the microwave network that runs transoceanic

I'm on the East Coast. If I put my microwave on medium and rig the door open, I can nuke a burrito from across the room. I figure if it was on high, I might just make it to London.

Earth isn't flat.

I also have a hand-held power drill, so a direct tunnel isn't out of the question.

Another consideration apart from my other response is cumulative flatness: Over any sufficiently short distance, let's call it X, that portion of the earth is indistinguishable from "flat" While the distance X[1] + X[2] might begin to exhibit a degree of curvature, you can take the distance between the midpoint of X[1] and X[2] and come up the same distance X, thereby bridging the very minor curvature of X[1] + X[2] with a span of flatness. In this way, any arbitrarily long distance can be treated as flat by chaining together the successive midpoints of X[1...N].

This is the theory of Non-Intuitive Cumulative Flatness (NICF) and is indeed rather ground breaking. The rugged persistence of flat-earthers over the years demonstrates that their claims must have some validity, and yet every last bit of empirical evidence refutes their viewpoint. NICF offers a unified theory of flat-curved earth geometry, much the same way that science has tried to bridge the gap between general relativity & quantum physics, to bridge these two contradictory theories and thus, demonstrate the earth to be both curved & flat.

Methods & applications that may take advantage of both geometries are currently underway. Most notably, finding the edge of the flat geometry would allow launching of space vehicles under significantly lower escape velocities, at which point the vehicle could transition to the curve geometry.

Softbank is highly interested, and a > $1billion funding round may be closing shortly after a successful MVP demonstration using a highly modded version of Kerbal Space Program.

I hope this is a joke, but I really don't know anymore.

every increasing tiering of pricing for different service models is 100% how I expect this story to go.

I think it might remain a pretty ok option for some, but this is definitely going to become a way to pay hella beans for your preferred non-neutral network providing.

Aren't the major HFTs in some cases just a few blocks away already?

Its for the people doing trades between exchanges (eg: New York to London) that this will benefit as they can act on the movement of another exchange before anyone else can. There is at least one microwave link between Chicago and New York to do this as the speed of light in air is faster than in glass.

HFT is arbitrage. The ask in Chicago falls below the bid in New York, so you buy for $12.25 in Chicago and sell for $12.27 in New York until there is either nobody willing to sell for $12.25 in Chicago or nobody willing to pay $12.27 in New York anymore. Which happens almost immediately, so whoever is fastest gets the money.

That means you want to be sitting right on both of the exchanges, but also that you want the fastest possible link between New York and Chicago.

How did you come to the estimate of 25%?

Because that difference in distance is the actual reason. The geostationary satellites used for your internet in 2009 are at ~2.5 of the earth diameter above ground (35k km, ca 1.5 earth circumference), whereas the starlink satellites are near the atmosphere (less than driving distance SF to LA). And speed of light is rather slow in these dimensions already.

I misread a different article which said spacex originally planned to have these satellites orbit around 800mi and I mistook that as normal altitude.

The analysis is basically the same for 800 miles. That's 5 milliseconds per leg. Geostationary is 22000 miles.

Though if you got 25%... did you calculate 115 milliseconds of speedup? That's the speedup for one leg. A ping to a server has to travel that distance four times.

Reply on news.ycombinator.com