Starlink Packet Routing | Hacker News Reader

Starlink Packet Routing

caseyhandmer.wordpress.com

84 pointssmithza36 comments

I would highly recommend the video linked from this article: https://www.youtube.com/watch?v=m05abdGSOxY

It's a simulation of the various ways in which the Starlink network could be arranged to route between satellites and ground stations; very detailed and super interesting.

This is kind of fascinating but it's also a fantasy that has nothing to do with reality. There's no reason why Starlink should hold itself to a ridiculously inefficient higher standard than other ISPs. Encrypt everything end-to-end and don't worry about ISP routing.

It also makes the big assumption of "when laser links are implemented". There is no timeframe for that, and it's possible that it's never implemented.

Already successfully tested (early September 2020).

https://www.teslarati.com/spacex-starlink-space-lasers-first...

Yup, and Starlink was also going to provide a 1Gbps service with a very low failure rate (now we know both of those aren't true). That article cites no sources, and SpaceX has not publicly said they've successfully tested cross-links. Even if they had, it only matters when they're actually being used.

What is the failure rate issue they are having? I haven't seen anything about that. Is this referring to their early satellites being decommissioned? Does that matter as long as their coverage stays the same since those were early prototypes and they are replacing them so quickly?

https://planet4589.org/space/stats/megacon/starbad.html

Ignoring the "0.9" satellites, this says 694 out of 713 launched so far are good. Is 97% success bad?

97% success is very bad when the number of satellite is so high that it causes space junk, yes. See here: https://twitter.com/ProfHughLewis/status/1312021242900295688

Thankfully these are all in LEO and the vast majority are under a controlled decay, and those that are not currently were most likely placed in a specific orbit to decay safely first per your starbad link.

The failure rate is low, everything else is just FUD spread by the competition who no they can't win in a fair fight. And how do you know 1Gbps is not achievable? Just because the first couple test publically available test didn't doesn't mean much really.

The have said that they have launched with laser links, however not that it was successfully tested them.

The existing satellite providers could provide a 1Gbps plan, but it's not economically feasible. Any of the multiple Gbps communications satellites could do that. SpaceX can also do it, but it's not economically feasible based on everything we know about the cost of the system. I know Elon gets a pass on pretty much anything he says, but it's not okay to continually make claims that will clearly not be true.

The difference is that Starlink will have 10000+ satellites and traditional systems have far, far, far, far less. Saying 'its not viable for <old system>' and concluding its not viable for Starlink is a terrible line of argument.

I would assume the base performance is not that much, but it clearly not some crazy number for the Starlink system.

The number of satellites has nothing to do with my hard the inter satellite links are. higher bandwidth links are much more expensive and harder to get right (harder meaning never been done at those rates). It's fine that you are in the camp of anything as possible, but that doesn't mean that it really is.

Often you don't even have to go threw inter satellite links, often it makes more sense to go down to the closes base stations. Specially if those base stations are right on the roof or next to the datacenters of the major companies. SpaceX base stations will likely be colocated with exchanges and that is also where many caches are (Netflix, Google and so on).

Even if you use inter satellite links, Starlink has many links and can schedule the path depending on demand on the network.

I'm not sure what you are even arguing, its very likely possible to get 1Gb, if SpaceX will over this, or only over it for limited time for extra price is clearly jet unknown. But saying 'current GEO satellites don't offer it so Starlink will not' simply doesn't work as an argument.

> harder meaning never been done at those rates

So nobody has done it so its not possible? That the worst argument ever. Laser link technology is at the very beginning. I think SpaceX had enough of 'experts' claiming things aren't possible even when physics doesn't at all show this to be the case.

No. If you have enough inter satellite links, it is almost always better to use those unless there is congestion. It's simple math.

You misread my other response. The reason I believe SpaceX will not offer 1G plans, except at a ridiculously high price, is because it's not profitable. GEO satellites could do it now if they wanted to, and their cost per bit is far lower than SpaceX.

Nobody has ever said physics is limiting this. Just because they were able to land a rocket doesn't magically make an entirely separate section of the company capable of doing things that haven't been done before. And the technical parts aren't even the biggest argument. The biggest argument is still that if they do not get government money, it is not financially feasible.

And I don't know what you are talking about with being co-located next to "exchanges", since that is extremely vague. The known POPs they are using thus far are not major ones. And because The satellites are so close to the Earth, they need many, many gateways in rural areas to keep the bandwidth high.

It seems unlikely that it would never be implemented. The benefits are pretty clear, and even though it isn't easy they're no fundamental roadblocks. Lasercomms has been done before, just not with hundreds/thousands of nodes in the network.

Sure, the benefits are very clear. But high-frequency inter-satellite links are very difficult. Facebook formed a group to test this early on, and eventually cancelled the project (and are now trying again). It's not just laser comms, it's the very high bandwidth needed in the last comms.

What about inter-planetary routing?

I'm guessing some of this lays ground work for having satellites between earth and mars, maybe some peering for batch updates or something.

Or do you mean something else by ISP routing?

You'd get a lot of timeouts doing interplanetary IP so there's Bundle Protocol: http://bundleprotocol.com/

I'm not sure I would say it has nothing to do with reality; that's up to SpaceX to decide. The premise suggested in the post, extremely briefly, is that, because Starlink would be a global ISP (which, to my knowledge, basically doesn't exist today), having much stronger encryption on the network would help them avoid needing to comply with a patchwork of differing legal frameworks across the world. Put simply, a truly global ISP exists in a fundamentally different regulatory and legal landscape than traditional ISPs, and the easiest route for them would be simply to sidestep them all by having provable anonymity.

I'm not sure how that premise fares in relation to the actual legal reality of operating a multi-jurisdictional ISP, but that's the premise, and I think it's plausible enough not to dismiss it out-of-hand.

What about jurisdictions with laws that you aren't allowed to intentionally design your system to obscure or forget information that may be relevant to a legal investigation?

Yeah they wouldn't be allowed to do this in lots of countries.

The only fundamental thing that’s different is that SpaceX doesn’t need to build the infrastructure in-country.

Subscriber relationships are going to be the same, and all of the regulations surrounding commercial operations are going to apply. If SpaceX can’t legally operate commercially in any locale then it seems incredibly unlikely that they will try. The satellites know where they are and their phased array antenna will know approximately where the terminal is, especially if gps is embedded or they do multilateration of a signal with multiple satellites.

This also is the big problem with the post, the satellites will be doing lots of beam steering to maximize throughout and minimize latency. Flooding the entire area underneath them with a single signal would severely impact service.

That said, they will almost certainly have to implement some or all of these controls in support of military operations.

I suggest checking out some of this blogger's other posts, or his book "How to Industrialize Mars: A Strategy for Self-Sufficiency".

Here's a good post on what's worth producing in situ, and what's worth importing from Earth over time.

https://caseyhandmer.wordpress.com/2020/08/23/progression-of...

How is billing supposed to work with this?

Are you asking if billing will be primarily/secondarily based on traffic use?

Cryptocurrency seems like a natural fit.

I’m half joking.

It should be fairly easy to come up with a cryptographic scheme that allows centralized billing, though metered billing may be more challenging.

“Every user terminal pointed at that satellite receives every data packet sent in that beam” Oh no no no, that sounds bad isn’t it, generally telcos avoid this problem by time and frequency modulation and providing a set of channels to each “terminal” where their data is sent and received, would that be possible for a fast moving “base station” in LEO? or is starlink a completely different system from my understanding.

Understood the writer meant electrical signal for receive. Does anyone know how they do latching and de registering of terminals at such speeds.

Cable modems and to some extent cell phones work the same way; that's why packets are encrypted with per-customer keys. You receive the whole beam but only decrypt packets destined for you. This is not a problem.

This is nonsense. First of all, it seems to assume that only Starlink is used for internet access. Ok, maybe we can go with that, even though in reality it would be extremely slow due to physical spectrum limitations. But why would packets be routed to a physical location? How could you get a network to work?

> I’m not an expert in networking

clearly. Also clearly, not an expert in privacy.

I am torned between cool new tech with lower latencies and are we really going to launch thousands of satellites polluting orbit in order to shave 10ms from connection?

Should we also stop using cargo ships because they pollute the water? Or planes because they pollute the air? Don't cables in the ground pollute the environment? Isn't every single way to get internet or anything to your house polluting?

These are controlled space craft flying in a known low orbit, they all have a life-cycle and will de-orbit once they are not needed and even if the propulsion system breaks, they will be on a known path and will deorbit within a few years.

Uncontrolled space craft and things to small to track are the real issue.

It's not 10ms, it's hundreds. GEO orbit ISPs provide a service that's complete garbage for desperate people with no other options. The latency is bad enough that voice and video calls are miserable, online gaming it out of the question, and just loading a web page takes 10+ seconds because of the round trips required.

Reply on news.ycombinator.com