in terms of raw capacity, I do not think that the intersatellite laser links (if/when they do get that working, which is a long ways off, 10 out of the 1050+ satellites right now have them, and none in production use), will approach anywhere near the capacity of a 40 or 80 channel, coherent 100/200/400GbE DWDM submarine line terminal. it's a whole other order of magnitudes.

Why couldn’t intersatellite laser links have similar capacity to fiber? In principle, if not practice?

Is it because ISLs will require more powerful lasers because the light is focused through a lens instead of guided through a fiber?

>> Why couldn’t intersatellite laser links have similar capacity to fiber? In principle, if not practice?

Because laser energies, signal paths, traveling through free space must literally share that space with each other. It is effectively one wire. An underground cable can have hundreds, thousands, even millions of wires lying beside each other without interference.

It is comparable to the difference between a wifi connection and bundle of ethernet cables. One can scale and one cannot.

I'll start by saying my understanding of the laser connection on the satellites is virtually nil.

Could you just add more beams on each satellite?

No. The beams would overlap, which is fine of they are on different frequencies, but eventually you run out of frequencies to use. Fiber confines the beam to inside the fiber. No overlap between one fiber and another means you can run multiple down the same pipe.

Why would they overlap?

Because a beam in an open medium always diverges due to wavefront errors and diffraction. A very good laser system might possibly have a divergence of only 1 microradians. For reference a typical laser pointer has a divergence angle of 1 to 2 milliradians. With the 2000km or so between satellites for Starlink the laser "spot" on the receiving satellite with the very good laser will be about 2 meter. Not enough to receive multiple beams on one satellite without overlapping.

Thanks for explaining. Would it work if they put receivers on a long pole, spaced 3m apart? Maybe they could mount them on the edge of the solar panel array, which is ~30m long [0].

[0] https://lilibots.blogspot.com/2020/04/starlink-satellite-dim...

The link estimates the surface of the panels to be 30m2. Since the panels are 3.1m wide the length would be 9.6 meters. But even at 30 meters it would be very very hard to engineer a system that can keep a laser beam aligned on a moving target with sub microradian precision. I'd guess Starlink laser are more divergent to make hitting the target easier. Keeps the exit pupil a manageable size too (narrower beams need bigger optics).

Ok, but I was asking about a single space laser vs a single fiber, not the theoretical limits of many of them.

Space is pretty big, lasers are pretty narrowly focused, and not many people have lasers in space yet.

High end fibre routers use multi-spectral bandwith, effectively allowing for multiple data streams to be transmitted on one fibre cable, and re-broken out at the other end. This significantly increases bandwidth.

I doubt the laser links support this, and I wonder if any could, when in a non-encapsulated environment. EG, spectral pollution from sunlight/etc.

probably less (literal) bandwidth, lasers are expensive

That and because the atmosphere is not clear (this is why 1Gbps free space laser network bridges have very limited range), the links from satellite to earth in any rf band will be very small in capacity as compared to a singlemode fiber cable. Space to space lasers, sure... Much clearer.

Starlink ground links are RF

That's what I said

oops my bad

Do the markets need lots of bandwidth? Or just-enough with the lowest possible latency

Markets don't, but, GP said:

"I can't imagine much appetite for any future under-sea cabling out of Australia"

And the reality is that satellites can not compete with the bandwidth of subsea cables for bulk data transfer, HFT is a whole other world.

I’d guess that high speed data transfer competes with overnight shipping a hard drive in most current cases. In my day to day life, for most cases I’d rather has 10 Mbps with 10 ms latency than 1 Gbps with 250 ms latency. Generally speaking I’m looking for a snappy user experience and fast enough for Netflix streaming to a non-premium television. But high bandwidth and low latency would be super nice!

A lot of serious HFT has moved to HF radio bands with giant aimed yagi uda or dipole antennas, already... London, new york, chicago, tokyo. It's very low bit rate but lower latency.

So Australian FinTech will definitely appreciate Starlink, then?

I suspect Australian FinTech would be more interested in colocating near exchanges instead of trying to beam their orders through space, but I don't work in the sector so I can't say for sure.

Totally different use cases.

If you want to observe events at one exchange and send orders to another, all that matters is total latency. For that use it doesn't matter whether your servers are in the same room as one of the exchanges or 200 miles away from either one.

It's true that total latency is what matters, and minimising latency by sitting near one of them and taking advantage of low-latency connections between exchanges sounds reasonable.

Does this have a name? Something like "a race to ridiculousness?" It's so bizarre that it matters that someone places an order a few ms before someone else. And it's probably ns now. No time to stop and think,... just be faster than someone else.

Capitalism? :P

Many wired leased lines that connect data vendors to exchanges are only ~2Mbps even in 2021.

You tend to get bandwidth spikes on most feeds at market open and close, but intraday rates are low

Check out this analysis of starlink ground relay routing vs fiber https://youtu.be/m05abdGSOxY

It's not exactly a long ways off if a dozen satellites are already using it (there were 2 previous ones).

Every time SpaceX makes a goal, someone claims it's really far away until it is completed.

If there was a need for Terabit/s laser capacity for Starlink, there's no particularly good reason it couldn't be developed.

Musk does tend to achieve his goals eventually although the timescales are generally 2-7 years too ambitious:

www.bloomberg.com/features/elon-musk-goals

Damn that doesn't bode well for me getting a Cybertruck any time soon.

Is there any mechanism by which we (in general) or a client specifically could use an expensive low latency for say browsing and a high latency but wide/cheap pipeline for bulk transfer?

I can imagine being able to do this when you control both sides by letting one type of data flow over one IP and the other over a different one with different peerings. But I wonder whether this is generally available.

In 1998, I lived in Ghana and used the Africa Online ISP. They ran a transparent HTTP proxy that routed through a high-bandwidth high-latency satellite connection. All other traffic went through a very low-bandwidth low-latency fiber connection. I used to proxy my browser through SSH to a US server and enjoy very fast text-mode browsing.

I expect some ISPs will end up using differential routing over Starlink and fiber. They could have the routers on each end automatically switch large one-way TCP transfers from the Starlink connection to the fiber connection after the first 2MB.

They could even make the transition smooth, by buffering data transferred over Starlink and slowly adding delay to match the delay of the fiber connection. This would require some RAM, but only during the transition period for each TCP session. The client would see the throughput go down during the transition period.

If the remote server could increase the data rate, then the ISP could perform the transition with no drop in throughput seen by the client. The sending router would send fake ACKs to the server to get it to send data faster. The sending router would send the burst of data through the fiber link and also buffer it and send it through Starlink at the original rate. Once the receiving router is receiving the same data at the same time over both the fiber link and Starlink, the sending router can stop sending over Starlink. The client's TCP ACK packets would continue to flow over Starlink.

If either your ISP exposes that ability or if you maintain multiple outbound connections, then yes. But I don't think this is commonly implemented today, however.

You would probably have to configure your own routing table to direct traffic for each application through the appropriate link, but that's completely doable as long as you can write rules to distinguish the packets (for example based on destination).

Yes I can imagine using iptables to work but it was my understanding that this would only be able to determine the next hop and I couldn’t get it to force the decision higher up the chain. This would mean that every party along the chain has to play ball.

I can imagine that it would be beneficial for the internet experience if we could segregate low latency traffic (human/hft) traffic from hulk traffic such as video. Although this goes against net neutrality I suppose.

If you want it to be done externally without setting up a routing table then the next best option would be some kind of traffic tagging to indicate what you want to go over which line. That would start looking like a network with multiple VLAN:s.

If you use TOR browser, you're already doing this. Requests from TOR browser go over TOR, if you have a regular browser open in another window your internet-routable requests there go over the internet. Of course the first hop goes via your ISP in either case, but it's easy to imagine it doesn't.

You can do this with a Linux box that has multiple internet links and/or VPNs, using fwmark for routing decisions. You can also e.g. launch a browser in a network namespace that is routed as you please.

Isn’t the current latency between Sydney and Singapore 15ms or less by now? No one who is latency sensitive should be using Starlink unless they are in some incredibly remote location. (And they should move if they are 150ms is a cosmic age in low latency trading)

Sydney to Singapore is 6300km as the crow flies. At light speed completely directly best possible round trip time would be 42ms. Light moves more slowly in glass(fibre) and the submarine cables are definitely nowhere direct route, add also routing/switching overhead and the you end up in the hundreds.

Just looked it up. 94ms. Starlink has far more latency.

By your lightsped math, Starlink won’t ever be able to compete on latency except in the most rural of areas with poor connectivity since low earth orbit has far more travel distance than convoluted surface routing.

Yup, we might be down to 100ms now -- certainly still a long way from your initial 15ms guess.

Since I moved on from that last gig, I note that Telstra completed their Sydney-Perth re-cabling project, and the Vocus Perth-Singapore undersea link came online (2018-04).

(I believe east coast traffic routes through Perth to get to Singapore, which was/is/will continue to be part of the problem.)

"lightspeed math" means LEO may well offer performance improvements over terrestrial cabling in many non-obvious scenarios.

Redo your math, in space the speed of light is at it's max. On fiber or cable the speed is cut almost in half. Result, Starlink with lasers can be faster.

Australia min latency from Singapore is still 125ms for me. Maybe it's lack of peering from my providers side but it's definitely an issue.

I get 75~ ms average on Aussie Broadband from Adelaide (FTTP) pinging Valve's Singapore game servers

It's ~50 from perth though.