How Just Four Satellites Could Provide Worldwide Internet (technologyreview.com) 108
We've known since the 1980s that you don't need mega-constellations comprising thousands of satellites to provide global internet coverage to the world. Continuous worldwide coverage is possible with a constellation of just four satellites placed at much higher altitudes. So why don't we have that? The big obstacle is cost. Several factors work to degrade a satellite's orbit, and to combat them, you need a huge amount of propellant on the satellite to consistently stabilize its orbit. Manufacturing, launch, and operational costs are just too high for the four-satellite trick. An anonymous reader writes: A new study proposes a counterintuitive approach that turns these degrading forces into ones that actually help keep these satellites in orbit. Instead of elliptical, the satellites' orbits would be circular, letting them get by with less fuel while still providing nearly global coverage (at slower speeds). The team ran simulations and found two that would work -- but there are still too many other issues for it to ever happen.
Slowly (Score:5, Insightful)
Slowly is the answer right?
Re:Slowly (Score:4, Informative)
Extremely so... not only is the latency going to be extreme due to the high altitudes, but having limited hardware (using what's essentially a shared medium) means that clients will be fighting for communication time. Having a high latency and high collision rate means you'll also have high retransmission rates, further increasing the demand for time.
Those conflicts can be mitigated somewhat with modern radio technology, but that raises the cost significantly for even modest performance improvements, and still leaves harsh limits on what a single multi-million-dollar satellite can handle.
Re: (Score:3)
Disadvantages:
1. Higher cost
2. Higher latency
3. Lower bandwidth
4. Less reliable / no redundancy
Advantages:
1. A more aesthetically pleasing orbit
Re:Slowly (Score:5, Informative)
5. Single point of failure
6. Failures result in days (moving a standby satellite into place) to years (building & launching a replacement) of service interruption
Re: (Score:2)
Yes, you just explained what #4 meant using more words. Good job.
Re: (Score:2)
Offtopic :
Please note that the post started with UID of 5 digits ( junta 36770 ) thats 1997 or 1998
then a 7 digit 2009-2012 Sarten-X
then a 6 digits 2003-5, Shanghaibill ( I've enjoy a few of this posters comments )
then a 7 digit 2009-2012 Sarten-X
yours 7 digit much more recent 2016-17 Lanthanide
Which I found funny and something to research.
the posting strings replies might be indicative of nothing
or
the education about documentation and/or explanation for the layperson as taught in schools
Since my hobby i
Re: (Score:1)
Re: (Score:3)
Also (I know it might be hard to grasp)
Re: (Score:2)
Re: (Score:2)
Ouch but a very factual statement.
Re: (Score:2)
It is, at least, a way to determine a *minimum* age.
Re: (Score:2)
Your point is valid, a Minimum age or as I use it mostly, a start point of tech sector views
Re: (Score:2)
I think it was around 2000 when I started reading here, but found the comments to be silly in-jokes and goatse... Certainly no reason to associate with this absurd community. I would read /. for the articles, I swear.
In early 2009, I was unemployed and found myself in rural Africa (which was surprisingly helpful to my budget) with Slashdot as my most useful resource to follow tech news. When I returned to the US, I made an account, and apparently posted my first comment on July 4th.
Make of that what you wil
Re: (Score:2)
It started *far* earlier than people forgetting their email.
One morning, we loaded slashdot and accounts were created--and logging in required *cookies*.
The reason my ID is so *high* is that it was a while before I had something I wanted to post enough to accept a cookie.
Back then, it was downright *common* to create a directory with the cookie file name to cause the writes to fail . . .
Anyway, anything into to mid four digits isn't time-distinguishable from two digits for this reason. Ultimately, the coll
Re: (Score:2)
Thank you for your response.
Sadly, I lost my 147K range username and id due to a company that failed and I could not get the email access I need to port it.
but otherwise it's always interesting to see what each generation of users post.
Believe it or not, there are from 1998 to 2018 about 5 fully different generation of programming styles and views towards tech ( and the self interest if tech ).
Right now, there is an interesting change of views, the /. generations from 1998-2006 in which it was information w
Re: (Score:2)
In hindsight, yep.
When I first read ShanghaiBill's post, I was still thinking about transmissions, and those having no redundancy, and remembering positioning a directional dish on a short building on the north side of a tall building... If the direct path's obscured, there's no other satellite to talk to.
Then that got me thinking about the issue of having a single satellite, and the problem with those being less reliable, and the orbital mechanics involved with maintaining a tiny constellation. Then I wrot
We already have such satellites (Score:5, Insightful)
"at much higher altitudes. So why don't we have that? The big obstacle is cost."
No. It's latency.
It's OK to say hello to meemaw, but that's about it.
Re: (Score:3)
Came here to say this. Yup, latency is THE reason.
Re: (Score:2, Interesting)
For 99% of internet usage latency is completely irrelevant. ... no one cares about latency.
The only usages where latency is important are games and remote surgery.
Emails, web browsing, chat apps
Re: (Score:2)
For some definition of 'latency'.
Latency of 200-400ms is acceptable for those applications. Once you get above those levels though, user experience quickly becomes frustrating.
Re: (Score:2)
Loading a random /. comment to answer to it, takes more than 400ms .... just like this one, I'm responding to, right now.
Re: (Score:2)
Having worked on cruise ships and supporting them remotely, the latency over satellite is simply awful. Typing over ssh, for example, is purely frustrating. Latency matters.
Re: (Score:1)
Re: (Score:2)
Appreciate the tips. Was aware of Mosh, although as the cruise customer controlled the rights for us to deploy changes or updates, we never got to use mosh beyond a one time test. :(
I did not, however, know about 'sam'. Thank you!
Re:We already have such satellites (Score:5, Informative)
This is an unbelievably naive statement, especially assuming you're someone that has used to Internet at least once. Just this /. page (with ad blockers) is loading two dozen different resources from at least a dozen different domains. There's another two dozen resources blocked by the ad blockers.
An uncached load of the page will take forever over a high latency connection. You're looking at half a second for the initial HTML load, another half second for replies to DNS queries for all the resources on CDNs and subdomains (assuming they all land at the same time). By default browsers don't make more than six concurrent HTTP 1.1 connections per server so six resources from any server will take half a second to load but a seventh will be a full second later. Depending on how the page is put together loading of new stylesheets or JavaScript can cause layout changes and redraws.
That's a non-TLS page load. A TLS handshake requires multiple round trips to the server (each with a half second latency). It takes even longer if a non-TLS endpoint redirects to a TLS endpoint. That needs to happen for each server contacts.
Some servers/CDNs will just cut you off if a connection has too high a latency as a high latency connection looks a lot like someone trying to hold open a connection as part of a DDoS.
All of that is assuming no packet loss which is never true; satellite connections have fairly high packet loss rates even compared to cellular connections to say nothing of landline connections.
If the only pages you ever viewed were singular documents on non-TLS endpoints on servers that didn't care about high latency connections...satellite can be workable. For everything else GEO satellites really suck for Internet access. Even streaming video sucks over satellite because any additional latency or packet loss beyond the physically imposed latency can affect playback. Most streaming services don't maintain super deep buffers unless you use their "download" options if they're even available.
Hey look data! [danluu.com]
Re: (Score:2)
Re: (Score:2)
Interesting point of view, nevertheless latency is for browsing the web irrelevant. You might have an initial cost, but then it is just the same regardless if you have high or low latency. E.g. clicking likes on FB or just browsing wikipedia.
Re: (Score:2)
This is correct, and why "satellite" internet almost always is augmented with a terrestrial low latency, low bandwidth parallel connection. DNS etc run over dialup or 3G and then the satellite handles the high bandwidth, high latency transmission.
Of course, for video transmission... we've already solved this 20 years ago with Tivo and satellite GEO video transmitters. Set your TIVO to download the low bandwidth TVGuide times and dates over dialup and then cache the movies locally as they're streamed over
Re: (Score:2)
Re:We already have such satellites (Score:5, Insightful)
This. We already have geostationary satellites that can provide service, albeit slowly, pretty much anywhere on earth. They aren't popular, but not because of the cost. HughesNet, for example, starts at $40 a month for service up to 25 Mbps. Rather, with geostationary satellites, the latency makes it borderline unusable unless you're just watching unidirectional streaming video all day.
For an Internet connection to feel responsive, you need double-digit milliseconds of latency. Even at just 100 milliseconds, your network connection feels slow, and some websites start to misbehave in weird ways.
The round-trip latency to a geostationary satellite is bounded by the speed of light. A signal must travel from the user to the satellite and back down to the ground station, and the response must follow that same path. So the minimum possible round-trip latency can be calculated by taking the satellite's altitude multiplying by 4 hops (up, down, up, down), and dividing by c (the speed of light).
Assuming you are at the equator (the shortest distance to geostationary orbit) and that the ground station is also at the equator (i.e. the signal is going straight up and straight back down), the minimum possible round-trip latency is 477 milliseconds (4 * 35,786km / c). If both the ground station and the user are in the continental U.S., that number increases by a couple of hundred km, i.e. the ground position is noise relative to the altitude.
The laws of physics will not allow you do reduce the latency to geostationary orbit at all, much less by the factor of ten that would be required to make geostationary satellites usable for Internet service. The only way to reduce the latency further is to move the satellite closer to the user and to the ground station servicing its traffic. That's why all of the planned Internet satellite systems use satellites in low earth orbit, at an altitude of about 2,000 km. At that altitude, the four-hop round-trip latency is only about 27 milliseconds, which results in service that is actually usable.
Re: (Score:2)
People like you may care but most of us, me included, have no trouble using Hughesnet. My service is $80 a month but it is infinitely better then nothing. My emails pop right up and YouTube works fine. The only thing I do that is slow is enter an HGTV contest every day. I can see dozens of handshakes and reloads as all the ads that are on the site try to load. The site will not work with an ad blocker so I have it turned off. With the ad block on all the other places I go work fine. Being old and retired I
Re: (Score:2)
I used to own a Ford, and before that, a Chevrolet. I stopped buying them because the number of serious problems I encountered was utterly insane.
On the Chevy, in addition to a number of weird electrical problems, some of which caused the car to not run at all, the paint also fell off in sheets.
On the Windstar, the rear electronics module went apes**t and unlocked and relocked my doors randomly while driving, an
Re: (Score:2)
I guess I am a lucky duck. I put 300,000 miles on my Aerostar, it was problem free for the first 200k. I had my 2000 Mercury Grand Marquis until last year when my wife got a Buick Encore. The Mercury was in perfect condition so I gave it to a disabled Vet who needed a car.
I am a good mechanic so maybe I can sense small problems and fix them before they get worse. Same thing with my computers, I keep them until they can't keep up with the increasing performance demands of the Internet. My last one was 10 ye
Re: (Score:2)
They aren't popular, but not because of the cost. HughesNet, for example, starts at $40 a month for service up to 25 Mbps.
I could not find the prices on the HughesNet site and other sources said $60 but assume it was $40 for their lowest cost plan. It only comes with 10GB so using the 95% provisioning rule common in industry, that comes out to about 38 kbits/second or slower than the dial up I used 20 years ago.
Re: (Score:2)
Eh. Draim's original 4-satellite system was designed with 8 hour elliptical orbits, meaning the mean orbital radius would be about 10700 miles, or about 6000 miles altitude, or around 65ms up and down. Not fantastic, but not horrible.
Re: (Score:2)
whups, math error. 8700 altitude, so 95ms up and down. Of course, that's at the mean radius, not the peak.
Re: (Score:3)
You're off by a factor of two. A round trip to a server via a satellite connection requires sending data from the user's dish up to the satellite and back down to a ground station, and then sending a response from the ground station up to the satellite and back down to the user. It's four hops, not two.
And 129 milliseconds is absolutely unacceptable latency. It's better than the almost half-second latency for a satellite in geostationary orbit, but not by enough to make it viable.
Re: (Score:2)
With the GP's update of the altitude to 8700 miles, make that 187 ms, which is almost half as bad as geostationary satellites, and almost an order of magnitude more than is acceptable, once you factor in the fact that you'll have all of the usual wired infrastructure latency added on top of that.
Re: (Score:2)
129ms latency is unacceptable for a land-based broadband system. It by no means makes the internet unusable. I used to play FPS games online with a latency of 150-200ms and it's perfectly doable, once you get used to it.
If you know that that is the latency of your internet connection then it quickly becomes 'normal'.
Only latencies of 400ms+ start to be unpleasant to use for web browsing, email, etc. But once again, if that's what you're used to, then you'll get used to it.
Re: (Score:1)
Re: (Score:1)
Re: (Score:3, Informative)
but I don't mind a 2 second latency for streaming a 2 hour movie...
of course games and web browsing are a different story.
Re: (Score:3)
Re: (Score:3)
Why say this? (Score:3)
It's not an extra quarter second, it's an extra quarter second per round trip. Most protocols are chatty, with several to many round trips for each interaction. Quarter seconds add up and become excruciating delays. No internet user with a reasonable alternative chooses a connection like this.
Re: (Score:3)
It's not an extra quarter second, it's an extra quarter second per round trip. Most protocols are chatty, with several to many round trips for each interaction. Quarter seconds add up and become excruciating delays. No internet user with a reasonable alternative chooses a connection like this.
If you want a perfect example of that phenomenon, just watch a local or national newscast where the on-site reporter is using a satellite connection to talk to the news anchor.
The noticeable delay between when one party stops talking and the other starts shows the issue perfectly. With the satellite in GEO (approx 22,236 miles out), it takes a little more time for the signal to make the round trip.
Latency plus the limited bandwidth of existing Internet satellites is a show-stopper for me.
But unfortunately
Reducing chatty protocols (Score:2)
We will need less-chatty protocols as we venture further into space.
There are some things like two-way person-to-person communications, real-time financial trading, and other things where noticeable latency is inherently bad.
For everything else, we can manage.
We already handle "one to many" communications like television or radio with multi-second latency and not even notice it most of the time. That "live" sports game is not really live, it's almost certainly going through at least one geostationary satel
Re: (Score:2)
NASA already has a 'deep space network' where they use less-chatty and error-correcting protocols.
Re: (Score:2)
A lot of us used to have to call relatives overseas back when geosynchronous satellite connections were the only way to call. Carrying on a conversation with that kind of delay was an exercise in frustration for everyone involved. Still better than no connection though.
Re: (Score:2)
It's an extra half second per round trip. To send traffic to a server, you have to send it up to the satellite, back down to a ground station, and then send the response from the ground station back up to the satellite and back down to the user. It is four hops, not two...
... unless, of course, the server is on the satellite itself.
And if you've ever actually tried to use an Internet connection with half a second of latency, let me tell you that it is anything but usable. Anyone arguing otherwise is jus
Re: (Score:2)
I don't think you'd even notice. You'd be too busy waiting because you're sharing a satellite with everyone else in your hemisphere.
Several obstacles (Score:2)
1. latency: GEO is a 0.7 s roundtrip.
2. cost of the terminals. You'd need a powerful transmitter or a large dish antenna to get a decent data rate for the uplink. IIRC TV satellites use transmitters in the 100 W - 1 kW range.
Let's not forget satellite internet from GEO is a reality. Services like Inmarsat and VSAT are used heavily by ships at sea and airliners. Those services cost too much to become ubiquitous though.
Re: (Score:2)
1. latency: GEO is a 0.7 s roundtrip.
As GEOnis about 1/10s of a light second "high", it is a bit higher than 0.2s
Re: (Score:3)
1. latency: GEO is a 0.7 s roundtrip. As GEOnis about 1/10s of a light second "high", it is a bit higher than 0.2s
You're forgetting that the server isn't in orbit. The request must go up to the satellite and back down to a ground station, and then the response must do the same. In total, it is just a bit less than half a second (4 * 35,786 km / c = 477 milliseconds).
Re: (Score:2)
Obviously we are not talking about severs, they don't care about GEO satellites anyway, the delay - latency - they introduce is always the same.
Re: (Score:2)
The point is that using a geostationary satellite for Internet service inherently introduces two satellite hops — one for a request, one for the response — so talking about a single hop up to a satellite and back to the ground is meaningless. It is equally true for almost anything else, including telephony. Thus, round-trip latency through a geostationary satellite is always just shy of half a second, at least for any meaningful use of the term "round trip".
The only situation in which a singl
Re: (Score:2)
The point is that using a geostationary satellite for Internet service inherently introduces two satellite hops â" one for a request, one for the response â" :D
Actually we call that one hop
Thus, round-trip latency through a geostationary satellite is always just shy of half a second, at least for any meaningful use of the term "round trip".
As I pointed out it is closer to 0.2 seconds than to 0.5 seconds, but that was just nitpicking and is not really important.
Re: (Score:2)
No, it isn't. .447 seconds is a lot closer to half a second than to 0.2 seconds. In fact, it's almost exactly half a second. Your definition of "round trip latency" is wrong.
Re: (Score:2)
First of all it is not 0.447 seconds, it is nearly exactly 0.2.
Secondly I did not use the "term roundtrip latency", you are using it. I only talk about the distance up and down, and is nearly exactly 1/10 of a light second, two ways and that makes it a little bit above 0.2 seconds travel time. Definitely below 0.3 :P
Thirdly nevertheless it is: one hop. And how long the router is processing it, no one knows.
Re: (Score:2)
From your comment [slashdot.org]:
That sure looks like you were saying that the round-trip latency is 0.2 seconds. The person you were replying to was clearly saying that the round-tri
Re: (Score:2)
Obviously I never talked about "roundtrip", so no idea why you want to nitpick about it.
Re: (Score:2)
Re: (Score:2)
Yeah, but if you actually do the math, that's almost lost in the noise, because geostationary orbit is just so ridiculously high. Nashville, TN, for example, is only 4,022.73 km to the equator, which translates into only a couple of hundred km difference in the hypotenuse of a right triangle from there to the equator. This is, of course, an approximation, because the planet isn't flat, but still, you're probably not talking about more than a single-digit percent or very low double-digit difference in dist
Bandwidth bottleneck (Score:1)
Re: (Score:2)
Re: (Score:1)
Latency kills, so does bandwidth (Score:4, Insightful)
So, lets see what happens when we replace 2B internet users onto a set of 4 satellites. Lets be polite and assume both the user population and bandwidth requirements are symmetrical, that is 500M users per satellite. Lets say each user needs 10Kbit stream (Note we say broadband is 3 orders of magnitude higher, but lets go with this and say that only 10% of people are on at any given time, and there is good bandwidth sharing - some people streaming netflix and some reading e-mail). That leaves us with a bandwidth of 5Tbit. That is an order of magnitude than the fastest Ethernet in existence today - and you would still have to ship the signal to space and back inside a power budget determined by solar and battery life.
Good luck with that one
Re: (Score:2)
Re: (Score:2)
Not sure why you're comparing to ethernet, when ethernet specifications are for 100m dinstance max.
Trans-ocean and trans-continental fibre backhaul cables are typically in the tens to hundreds of terrabits.
Re: (Score:2)
is that 100m for meters or miles. ...
because back in the late 80's and early 90's I ran 10base5 fat cables way over 1300 feet ( ran spools sometimes to the end LOL )
Have not done anything with cables in a long time so I can just imagine the distance of what a fiber-optic version of Ethernet can do.
Re: (Score:1)
Re: (Score:1)
Re: (Score:1)
Dumb (Score:3)
This is dumb. The ability to geographically cover the planet with service is not the same as having the bandwidth to make it useful. Unless this is intended as a guide for billionaires to put up their own satellites like private wifi hotspots in space.
Huh? (Score:1)
For ever ... (Score:2)
A satellite in a circular orbit at GSO or beyond will stay there for ever ... no fuel required.
Re:For ever ... (Score:4, Informative)
A satellite in GSO may stay in orbit for a very long time, however fuel is needed for station keeping as
you generally want/need to keep it "fixed" in location above the Earth.
https://en.wikipedia.org/wiki/... [wikipedia.org]
Re: (Score:2)
There's some AA-battery things in GSO designed to broadcast "I'm here... this spot isn't being used by anybody!" If anybody wants those spots they can be blown away.
Total bps (Score:2)
Re: (Score:2)
Yep, and the connection would suck (Score:2)
I worked at an ISP in the 90's, we ran a T1 and when that got full we ran a download link over a satellite connection. The ping times were in the 700ms range on a good day. That is restricted due to the speed of light, and for most geosync orbits it's going to be in the half second range. We already have satellites up there that can provide internet connections. There are a few reasons why everyone isn't using them.
1) Cost
2) Bandwidth, your restricted by radio bandwidth
3) Latency
A cell tower can handle hund
You only need 1 big sat! (Score:2)
The world is flat remember? Land lines are plenty fast for spreading conspiracy videos on youtube.
Re: (Score:2)
Even the ancient Greeks knew the world was round, and had a pretty good estimate of its size.
Re: (Score:2)
If you think the world is flat, you've probably been spending too much time on youtube.
It's about latency, not speed (Score:2)
Modern users are less interested in speed than they are latency. They don't really need much higher bandwidth than is available now, but lower latency opens up whole new realms of technological applications. Satellite doesn't cut it.
You can't be Sirius (Score:2)
Uh, the Sirius satellite system didn't work very well. It was designed to be Low Earth Orbit, but they kept hitting other things in the sky. Now, Sirius-only receivers are obsolete, and all North American satellite radio users use XM's space points.
Re: (Score:2)
Wrong! (Score:2)
The advantage of having LOTS of satellites in LOW orbits - rather than 3 or four in GEO - is that latency is minimized. Satellites in GEO have nearly 2/3 second of latency, from the signal having two round trips to GEO. Gamers, fro example, find that intolerable.
Bandwidth? (Score:1)
"the satellites' orbits would be circular" (Score:2)
Did Kepler's first law of planetary motion just get flushed down the toilet?
Four? (Score:2)
Arthur C Clarke did it with three. That is, until Big Satellite took him out.
Latency....unavoidable. (Score:2)
Sure, you can theoretically do it with four points at the vertices of a tetrahedron - and 100% of the surface is visible 24/7.
But they have to be so far away - that speed-of-light constraints kill you.
The ENTIRE point of StarLink is to keep the altitudes as low as possible to minimise latency.
I can't wait! (Score:2)
To play Google Stadia on a connection with .7 to 1.5 second ping.
Been tried (Score:2)
Commercial geostationary communications platforms have been tried, it didn't go well. The problem (from my limited understanding) is that a limited number of large satellites also have a limited bandwidth capability. You can only fit so many transmitters/receivers on a single satellite, and you can only fit so many solar panels to power those transmitter/receivers. So that limits your customers which in turn increases the pricing which in turn limits your customers even more, its a nasty feedback cycle.
Sirius Satellite Radio tried this, and it failed (Score:2)
Sirius Satellite Radio tried this, and it failed.
Their highly elliptical orbits worked well for reception in cars, but failed miserably for stationary reception.
Their highly elliptical satellites were decommissioned years ago in favor of geostationary satellites.
a few problems (Score:1)
The last sentence of the article says it all (Score:2)
"In the real world, engineering and economic obstacles might dim hopes for an easy solution."
So the study/article is just a brain storm.
Is Four Enough? (Score:2)
Yeah, you can probably cover the globe with four, but would that be acceptable? How long would it take to replace/repair one if it broke down and a quarter of the planet was suddenly going without? I can't imagine not having some form of backup coverage by at least a 2nd and likely 3rd or more. Especially, now that you've got countries building anti-sat weapons.
Re: (Score:2)
Inmarsat developed the Global Xpress platform, launched in 2015, which provides 50 megabit downlink and 5 megabit uplink worldwide with four satellites.
The existing satellites use hundreds of spot beams and the next generation of birds will use thousands of spot beams.
This has already been achieved with 4 satellites (Score:2)
This has already been achieved with four satellites.
Inmarsat developed the Global Xpress platform, launched in 2015, which provides 50 megabit downlink and 5 megabit uplink worldwide with four satellites.
Not sure what all the fuss is about.