The cutouts vary. It's not "every 10 minutes". DSL latency is not standard, it can vary a lot. Latency for the beta is worse than it will be for production. For the people this is marketed to, it's life changing. If you think it's overpriced, you were never the target demographic.
Almost. It's possible for a company to overprice the product for their target demographic. But then it's likely that they fail to acquire customers and go out if business. So it's generally true.
The article specifically has a quote from someone in Montana on CenturyLink DSL paying $120/month for 1Mbps, and getting 100x faster connectivity through their $99/month StarLink service (with some intermittency because the constellation and ground station network isn’t complete). Latency is acceptable enough to play a first person shooter, and target latency is 16ms-19ms (and meets the requirement for the FCCs rural broadband contract [1], which is latency below 100ms).
For those with marginal service, this is a material improvement. It’s also firming up the case that StarLink will generate billions of dollars per year in revenue (eclipsing SpaceX’s launch revenues), from both government contracts (rural broadband and DoD) and unsubsidized subscribers. This doesn’t include revenue from StarLink’s contract building missile tracking satellites for the Defense dept (based on StarLink sats) [2].
That's wrong. In this case, as with the iPod, the tech enables a dramatically better user experience. Compared to existing satellite ISPs you get dramatically better latency and dramatically faster/easier setup (no professional installation required, plug and play in 5 minutes). It's also likely to be far cheaper if you actually transfer a decent amount of data, and early tests seem to show better reliability in poor weather as well.
It's even better than DSL and some wireless ISPs or 4G hotspots with marginal connections, which are probably the main competition for Starlink.
> In this case, as with the iPod, the tech enables a dramatically better user experience.
The whole reason CowboyNeal was attacking the iPod was that technically it was a step backwards. It couldn't do anything that a Nomad couldn't. The part that it did well (and that he missed the importance of) was the usability and polish.
> Compared to existing satellite ISPs you get dramatically better latency and dramatically faster/easier setup (no professional installation required, plug and play in 5 minutes). It's also likely to be far cheaper if you actually transfer a decent amount of data, and early tests seem to show better reliability in poor weather as well.
All true, but can you see HN caring about any of that if SpaceX wasn't involved? The HN userbase is overwhelmingly city folk; I suspect only a tiny proportion of us have ever used satellite internet before.
Yes, I can see HN getting excited about a dramatically better satellite internet option regardless of who it came from. OneWeb, Amazon, etc could have had this attention if they had gotten here first.
Um, no — people are hyped because it will provide decent low-latency connectivity to areas that can’t even get a DSL line or more than a couple of bars of LTE.
the beta test period will probably have a higher ratio of uplink to user bandwidth than any time in the future.
these satellites are radically undersubscribed right now, serving a handful of users tops. that's not going to model what happens if this works & grows.
I think it's disingenuous of ArsTechnica to report two of the worst case posts on the Starlink subreddit as if they were common.
The source used for the comment "every 5-10 minutes" doesn't fully clarify that point. Many users in the Reddit subreddit post that is quoted stated that the person likely installed it incorrectly and the cutouts were from trees blocking the signal. So "every 5-10 minutes" is not what most people are seeing. Source reddit post the article uses: https://www.reddit.com/r/Starlink/comments/jsnd52/starlink_d...
Key reply:
> Based on the disconnects every 5-10 minutes. I would definitely say the obstructions are causing issues. I had the same problem last week and timing between disconnects. Put mine a little too close to the house. Moved it yesterday and today my connections were holding with no disconnects for anywhere from 2 to 3 hours with maybe a short 10-30 second disconnect between.
Given the overall upbeat tone of headline, subheading, and article, is "disingenuous" the word you want to use? Is it your impression that Ars is attempting to deceive or trick its readers?
Something feels very wrong about covering the entire planet in satellites just so rural residents of the developed world can get faster internet.
Sure, in theory, the entire world can have access but that won't come for free and won't be in any way affordable by the vast majority of the world population.
Indeed, combine a SpaceX backhaul with a 4G base station and you can provide essential service to even extremely remote or rural communities at an incredibly reasonable price.
Throw some solar panels and batteries on there and you can have portable self-contained 4g base station that you can drop off anywhere covered by starlink and sunlight.
There was a Native American tribe living in a remote area that got early access to Starlink. Now their children can do remote learning and everyone can get remote healthcare.
Seems like it would be an easy thing to set up with various foreign aid programs, as well. Sponsor the internet for a village, akin to getting a foster child or similar
> won't be in any way affordable by the vast majority of the world population
I think you have a pretty misguided view of the developing world and their buying power. There are a lot of middle-income countries with reasonable spending power, but lousy infrastructure, in-between the extremes of the undeveloped Congo (which yes, will take some time to get Starlink penetration) and residents of a NYC condo.
Why wouldn't it be affordable? As long as there's enough bandwidth couldn't SpaceX just offer its ISP services at marginal cost? In any case it's not like people who can't afford it are made worse off by it being theoretically available to them
Do you think the non-developed world only consists of dirt poor peasants? There are lots of people in these countries who this service would help. What you also seem to ignore is that many business in these countries would be helped by this.
Another thing you ignore is that if you combine Starlink with a mobile station you can potentially provide internet to a whole village.
Also, the prices will likely not be the same everywhere. SpaceX will want to get higher utility of the overall system and having unused sats fly over Africa is pointless. Lowering the price in some regions to get full utilization if an economic non-brainier if they would actually have issue selling the service.
In addition, the idea that this service wouldn't be worth it for rural people in the first world is very questionable. Don't rural people deserve good internet?
Sounds pretty awesome! Anyone know how much performance might be expected to degrade as the number of users increases? Most of these people probably have the entire Starlink RF spectrum to themselves in whatever location they’re testing in.
The problem will not be the spectrum, the sats will likely be the bottle neck. The amount of users can scale with the amount of sats you have.
You will hopefully get a better signal by having many sats in your view and the ability for the network to make smart choices about where to send what when.
The current Starlink launch is limited to the US and parts of Canada. What would happen if I sign up there and then physically ship my dish to Saudi Arabia/China/North Korea? Would the service keep working (subject to satellite availability, which will obviously not be great), or are there geoblocks in place?
The terms of service you agree to at signup dictate you’ll use the equipment at the service location you provided with your order, and the constellation is able to determine ground station positions (StarLink can, theoretically, replace the functionality of the GPS constellation using ~0.8% of total downlink capacity). SpaceX would likely comply with their FCC fixed station license and US gov sanctions requirements through technical mechanisms.
In addition to the points made by toomuchtodo, I believe Starlink currently relies on the presence of nearby ground stations to relay your packets to/from the Internet backbone. Eventually they plan satellite-to-satellite hops but I don't think that's enabled yet.
It'll be interesting to see whether SpaceX allows, relatively easily and simply, Starlink to be used by brave people living under oppressive regimes: it could allow for far freer information than what's currently available.
For now, but only because SpaceX is playing nice. If SpaceX decides to transmit to rural Syria against the wishes of the Assad regime, and the people on the ground find a receiver, who's going to stop them*?
Interestingly enough, the extent of a country's sovereignty in (air)space is legally quite unclear. It's pretty well established that countries can regulate planes and such flying at an altitude of ~10 km, and the Outer Space Treaty bans claims on celestrial bodies, but everything in between (including Starlink's satellites in LEO, some 2,000 km from the Earth's surface) is in a grey zone.
Legal issues aside, they're also quite literally beyond the reach of all but a small handful of spacefaring nations, and even a nation like China that has the technical capability would not resort to actively shooting them down unless other avenues were exhausted.
Some beta testers have moved around their antennas and they worked even when not at the 'official' location. So currently it 'seem' like they are just relaying on beta tester to be mostly honest.
Once you have real sat-to-sat laser comms (next generation of sats, currently being tested) you can use it anywhere theoretically.
We don't know what the long term strategy is around geo-blocking. In theory they could have an issue where some location is fully subscribed and then somebody is moving there causing an issue. It will be interesting to see what policy they will use.
The thing with satellite internet (up and downlink) is mostly to do with jitter. You can get pretty good rates on a nice day, but you can’t control air refraction 2 miles off the ground doing weird stuff to your signal. I experimented with many things, but the best solution I had was using TCP over UDP with a server in a data center able to resend TCP (over UDP) packets if they were dropped. I think http3 will help out immensely here.
Source: I ran a small satellite ISP on a base in Afghanistan.
He probably saw benefits from using UDP because he's no longer limited by TCP's congestion control and can flood the pipe with duplicates of his packets.
> He probably saw benefits from using UDP because he's no longer limited by TCP's congestion control and can flood the pipe with duplicates of his packets.
But he wrote that he'd used "TCP over UDP". The first lines of the TCP over UDP spec already state
"It provides exactly the same congestion control, flow control, reliability, and extension mechanisms as offered by TCP." [1]
Yes, I had the same confusion. There seems to be a standard for TCP over UDP [1]. But in general, I would assume that it does not help at all WRT any of the packet loss/latency problems posed by satellite links.
TCP is designed for encapsulation in IP, this is why it is called TCP/IP. And UDP is very similar to IP, only difference is that UDP gives you source and destination port numbers and a payload checksum, where IP only gives you a protocol numbers and non-checksummed data. None of these "features" is going to have any effect on how TCP is delivered.
So the only effect that use of TCP over UDP would have here is:
- increase the overhead used for headers, thereby decreasing the MSS (segment size)
- "cloak" the TCP traffic as UDP and thereby prevent any routers and/or middleboxes [2] from applying special treatment to TCP (i.e. traffic shaping, buffering etc.)
Maybe the latter effect is what made a difference for the link in question.
TCP does have retransmission, but it was designed for networks with very low packet loss and where most packet loss is due to congestion, so it does not react quickly to packet loss and when multiple packets are lost it slows down dramatically. TCP typically uses an exponential backoff in re-transmission times to avoid exacerbating congestion.
When dealing directly with UDP, one can implement different application-level logic for dealing with packet loss: "NAK" packets, different backoff strategies, and even forward error correction.
Sorry but you're wrong, dedicated 1:1 capacity scpc geostationary links will always be 492-495ms rtt ping latency between teleport and far end. Latency will never be less than the laws of physics and modem FEC dictate, but jitter will be very low.
If you saw a lot of jitter it was because you were in an oversubscribed, contended tdma network. What was your terminal hardware, what modems were you using and what was the rf chain and modem setup at the far end (not in Afghanistan)?
Latency on terrestrial ptp microwave also does not vary. Ducting and diffraction and temperature inversions on long 6GHz links affects rsl and modulation, not latency.
In the case of starlink, each satellite is functioning as a bent pipe relay between cpe and starlink earth station. Latency will vary a ms or two based on satellite position during a pass overhead, but no more than that. The rest of the latency is the same as from any terrestrial internet customer in the same geographic location as the earth station. For instance one of the beta test earth stations is in north bend, wa, colocated with a level3 long haul dwdm site.
People in WA have been posting tests with latencies of 25-26ms to speed test servers in downtown Seattle. That's actually really, really good, considering that latency to your default gateway on a residential cable docsis3 coaxial connection can easily be 16ms on a terrestrial connection. North bend is a few ms from seattle, the brewster starlink teleport site several ms more.
In the case of starlink they are competing against 32:1 or worse oversubscribed geostationary consumer grade vsat services like viasat and Hughesnet. Starlink is far superior tech.
Unfortunately I do not, I learned on the job. I would recommend learning general network engineering and study radio, microwave, physics and wireless communication stuff generally to have a good basis to begin.
For the RF side of things a ham radio license is a fun thing to have. Although you do have to dig a bit to find sections of the hobby that focus more on the digital side of things.
Very interesting. We once had a microwave link setup that would fail during the rain. Admittedly this was maybe 20 years ago so maybe the tech is different, but this would manifest as latency spikes presumably as there was a failure to extract signal and TCP backoff kicked in. i.e. it wasn't physical layer latency but transport layer latency.
But my knowledge of networks is fairly rudimentary and old now and now that I think about it, pings would get sudden spikes in the rain too and that's just a lower layer than I'm talking about. hmm. I was also a child then so I might be misremembering.
The modern tech is indeed quite different, a modern microwave ptp radio system in rain susceptible bands (11,18,23GHz, etc) will be capable of adaptive coding and modulation to change to looser modulations on the fly during rain fade. With no change in latency, but drops in aggregate Mbps capacity. Of course no amount of ACM will help if someone has engineered the link wrong, at a distance too far, antennas too small, or similar.
For an example of a modern radio google aviat wtm4200.
Interesting. They appear to be much cheaper (under $15k for the 4100 for the radio). Very cool very cool. Funny how in my head I've been judging a modern Monkeybrains installation by my 20 year old experience.
It's not at the same level but Ubiquiti makes some impressively "cheap" gear in this space that will do 1gbps over decent distances[1]. I've mostly used their lower powered p2p links but it's amazing to see how far the tech has come.
I would caution people against using the af11 or AF24 if they don't have to. Ubiquiti doesn't know how to make a serious radio.
For something like the af5xhd, it's a great little thing for its price, and with good quality antennas at both ends can make a very effective bridge. But for more serious things? No. It's so bad they haven't even been able to implement snmpv2 on their $1000+ radios. The af11 is so bad it only uses a 56MHz channel mask in an 80MHz channel.
No matter what manufacturer, serious ptp radios can be distinguished by two things: the presence of terminals for a direct -48vdc wiring connection and sfp/sfp+ ports.
At the very low end of cost, ubnt's 802.11ac chipset ptp bridge radios are great and very cost effective, I'm talking about the powerbeam ac series, nanobeam and similar. For when you want to buy a few hundred bucks of stuff on your credit card to link a few buildings together quickly. For anyone looking at that category of stuff they should also look at mikrotik and mimosa radios.
I like that some random guy on the internet is telling a guy who was in charge of the satellite internet on a military base that he doesn't know anything about satellite internet.
Lost packets increase latency. And weather conditions cause lost packets.
> I like that some random guy on the internet is telling a guy who was in charge of the satellite internet on a military base that he doesn't know anything about satellite internet.
More like everywhere on the internet. Just the other day, I saw some idiot on facebook try to lecture a board certified pulmonologist at one of the state's largest hospital about how covid is nothing to worry about.
If you want to get pedantic, the band used will determine what natural phenomenon impacts the received signal. Even in those cases FEC and dynamic modulation will let the signal work(at reduced throughout).
That said unless we're talking about lasers air refraction doesn't usually impact RF(there is tropospheric ducting but that's something different).
In an example of a modern highly rain fade resistant ptp microwave system in the 18GHz band, standard fcc part 101 fdd band plan. The radios might be talking to each other in both H and V polarities, 80MHz wide channel in both, at 2048QAM 5/6 during a normal sunny day. In an extreme rain fade event they may go all the way down to QPSK 1/2 and the link's capacity in Mbps could be as poor as 40Mbps (from what is nominally a 900Mbps+ link), but the end to end latency will remain pretty much the same.
Latency as experienced by users might be much worse, however, if suddenly 300Mbps of traffic is trying to be shoved through a 40-75Mbps link, for a period of several minutes or more.
uhhh packet loss doesn't change latency? Its loss, which has its own effects, but jesus dude you're just doubling down on you not knowing what you're talking about.
Packet loss on an oversubscribed tdma, contended, shared access cheap VSAT system is a result of too much traffic and not enough pipe. For an analogy think of a big office sharing a 5Mbps DSL line. A properly implemented dedicated geostationary link doesn't have packet loss except under sun transit events, extreme rain fade (for ku and higher bands), and when too many people try to stuff too much traffic through too narrow of a link without appropriate QoS.
From the point of view of an end user on a LAN at a site dependent on two way satellite it may not be possible for them to determine exact why they're seeing packet loss.
We're taking about two different types of latency. You, and the other people telling me that I'm wrong, are referring to latency at the packet level, and I am referring to latency at the message\data level.
https://en.m.wikipedia.org/wiki/Latency_(engineering)
Ultimately, the latter is the kind of latency the user cares about.
Credentials are not that important here. Maybe the random guy also ran satellite internet on a military base, maybe they manufacture the satellite comms gear? The important bit is, are either of those supported by facts and where the technical difference / misunderstanding comes from? (I wouldn't start with "you're wrong" but that's a separate thing)
They can both be right as walrus01 seems to talk about latency of a single rtt in isolation, and withinboredom about what does the TCP endpoint effectively see.
What I'm talking about is the nature of a dedicated access (1:1) single carrier per channel geostationary link, via a pair of modems such as comtech cdm760 and appropriately sized rf chains and antennas on both ends. It'll be a rock solid 492ms and the bandwidth (in Mbps) will be proportional to the size of the transponder space leased, and influenced by antenna sizing/gain, overall link budget, tx power on both ends and so forth.
Example: at one location, a 3.8 meter ku band dual linear polarity compact cassegrain dish in somalia, a 200W BUC+SSPA, good quality norsat PLL LNB, a cdm760 modem. Ten MHz of transponder space occupied total. Other end, a 5.5m dish with similar gear on it in a teleport in Cyprus.
If the person I'm replying to isn't willing or able to clarify whether they were in a contended, or dedicated network, or specify the gear, it's hard to say exactly why they're seeing jitter. But I can quite confidently say that excursions above 495ms are not something seen in a dedicated network.
It's how that dedicated capacity can get sliced up and shared that is the most common cause of packet loss, jitter and high latency in geostationary based two way satellite IP links.
The most common thing for a cheap VSAT isp to do is to acquire one chunk of Ku transponder capacity (let's say about $8000 a month worth), install something like an idirect hub with line cards in it, and sell access to individual end users in $developingnation at an 8:1 oversubscription ratio or more. Meaning that an end user vsat terminal might see 495ms at a lucky time when its peers in the same network were not also moving traffic, but could also often see latency anywhere from 500 to 850 during busy periods.
Now that same 8000 bucks a month of transponder space could equally well be assigned to a single high priority end user with money to pay, in which case they would never see above 495ms. Or it could be oversubscribed to all hell, such as if the ISP wanted to sell $280/month service to a bunch of very cost conscious end users in a developing nation environment who can only afford very cheap VSAT terminal hardware (one 1.2m fiberglass compression molded reflector, a 4W BUC, and an idirect modem) and minimal monthly ongoing payments. In which case each user will be in a 16:1 or worse oversubscribed network.
The economics of making a 5500 kilogram $200 million satellite in geostationary orbit pay for itself, before its lifespan is done in 15 years, are not favorable. If you ever want to truly be shocked go price the $ per Mbps per month that results from the 1:1 dedicated setup I described above.
On the topic of what the average hn reader considers "satellite internet". Based on consumer experience with the absolute cheapest consumer grade ku and Ka band vsat terminals in the usa, they almost 99% of the time mean Hughesnet or viasat/wildblue. Very few people actually have money or reason to lay their hands on real two way satellite stuff. Generally if you work for a company that needs something like a $3500/month link for a mine in Nunavut you're paying some third party experts to implement it.
Note that may be true for one bird in Afghanistan, but the path length isn't exactly the same everywhere, nor are the delays of all modulaiton schemes.
It will vary a few ms but not much. First hand experience, latency differs at most a millisecond or two between an Afghanistan-germany link, and an Afghanistan-singapore link. More latency variation to internet based things is seen based on where the teleport is. The same as would be experienced by a terrestrial internet user in either Raising, Germany or near the Singtel teleport in Singapore.
For instance our site in Germany was more than 15ms away by fiber from DE-CIX stuff in Frankfurt. The site in Singapore had better latency to any cdn or hosting operation that was in Singapore itself, but much worse if traffic had to find its way through the terrestrial submarine fiber internet to other places. End to end latency between Afghanistan-germany-seattle or Afghanistan-singapore-seattle was considerably different.
Yes, the position of the two end point sites on the sphere that is the earth will have slightly different speed of light times to a satellite in geostationary orbit. But 99% of that latency in the above 490ms figure is the extreme distance of the satellite, and not whether or not one site is slightly further down the sphere, as viewed from a theoretical observer sitting on the satellite.
It is also worth noting that modern scpc/mcpc modems can be intentionally configured to use more computationally "expensive" FECs, in a trade off of slightly more bps/Hz squeezed into a given sized slice of transport kHz, at the expense of a few ms higher latency. This is a network engineering design decision made when a link is first brought online. Modern modems and ASICs, FPGAs have made this less difficult than it used to be. If I can use an analogy for software developers, this is sort of like the speed vs file size trade off choice made between using gzip, bzip2, xzip, at different settings and compression levels.
The slant range has a significant effect on latency. For two ground stations at the edges of the satellites footprint, each with 10 degree antenna elevation, the latency is around 540 milliseconds.
> But 99% of that latency in the above 490ms figure is the extreme distance of the satellite, and not whether or not one site is slightly further down the sphere, as viewed from a theoretical observer sitting on the satellite.
Path length varies from ~34500km to ~42000km (~115ms to ~140ms). So, assuming an ideally placed teleport, from 115 * 4 = 460ms to 115 * 2 + 140 * 2 = 510ms speed of light delay on round trips. This is why I said I found 492ms to be an amazingly over-precise number. Just call it 500ms. :P
(I'm definitely not very knowledgable about any of this, naive questions incoming.)
> 3.8 meter ku band dual linear polarity compact cassegrain dish
> a 5.5m dish
Are these receiver sizes in the range of what Starlink expects from users? I had in mind something more like a DirectTV dish. If the consumer dish sizes are indeed smaller, how does that affect the discussion about jitter and other aspects of link quality?
How does that affect the discussion about jitter and other aspects of link quality?
With only a tiny bit of the relevant subjects under my belt, I would expect the metrics to be significantly worse using a much smaller dish. But it seems like you're suggesting Starlink could achieve similar metrics on a smaller dish? I'm curious how that works. Apologies if I'm misunderstanding.
What does the dedicated case have to do with starlink? It will be completely oversubscribed once they are out of beta, or they simply don't make money. So the original comment stands in that latency will vary.
Wow. I wrote that comment, went to sleep, then work, and now sitting down and eating my lunch... didn't expect to see this blow up. :)
Let me see if I can answer some of this.
We were on a 1:1, eventually. Tried the 8:1 (or something like that) at first and it was terrible.
Latency was almost always, like you said, almost exactly 495ms at the link. However, we'd suffer transient jitter that directly correlated to weather in the atmosphere. Essentially, satellite communication is a laser (parabolas ftw), so this made sense at the time. Correlation doesn't necessarily mean causation, but it was annoyingly correlated.
I only made this correlation because my "day job" in the military had a lot to do with radar, microwave antennas, and optics. So, I'd observe the refraction personally (being able to see mountains as clutter for example, when the radar rays were pointed well over the mountains) and then notice the observed jitter on the network. Not to mention people coming to tell me how unusable Skype was at the moment and could I fix it please?
No offense, but you probably don't see this at your level as a network engineer unless you're "boots on the ground."
When you're there and easily accessible, people will come complain to you when any and every little thing that goes wrong and "their wife is threatening to leave them and they can't even have a conversation, omg, fix it right now plz."
In a more corporate setting, people are far less likely to complain about transient issues because they fix themselves before the complaint can be made, or if they do, it's easy to dismiss as other issues.
To be honest, we had those other issues as well. QoS was a thing, and constantly battling bittorrent clients and their users was also a thing. But those issues were usually only during the evenings when lots of people were trying to call home and get on Facebook. I'm talking about when none of those issues were at play, such as late at night or during the day.
The UDP solution was directly related to periods of saturation, specifically I built a custom transparent proxy forked from Squid. It's sole job was to act as a mediator between a server back in the US that did a few things:
1. On the Afghanistan side, it would wrap the HTTP(S) request packet and send it via a TCP control channel to the server in the US.
2. On the US side, it would do some image compression (if applicable) and other things to the response to make it smaller (in bytes).
3. On the US side, it would also keep track of user flows and how much bandwidth each user got, and spit back UDP packets, reserving a dynamic % of bandwidth for control and unregulated packets. There was also an open TCP control connection for sending back smaller things that fit within a few packets.
4. On the Afghanistan side, it would put all these packets together to reform the stream, keeping track of any dropped packets (rare) and then forwarding the completed stream (or once the buffer was full) to the user at 100mbps. If any were dropped, it sent back a bitmap over the open TCP connection to have those frames retransmitted via TCP and then inserted into the stream before going to the client.
It worked fantastic for things like youtube, since the subscriber could make full use of their subscribed bandwidth with a half-second delay after clicking a link. The only things that didn't use that system were things like Skype. We never really had transport-related congestion issues after this was built as a hack, over several weeks. To be honest, it probably would have worked just fine by forwarding over already open TCP connections to the back-end server and giving each subscriber a dedicated stream. However, time in the desert is a premium and once something works good-enough, it pretty much stays that way forever.
> If the person I'm replying to isn't willing or able to clarify whether they were in a contended, or dedicated network, or specify the gear, it'...
So you had physical problems with the signal, and you had jitter, but the physical problems don't directly or inevitably cause jitter. The cause of the jitter has to have been a flawed method of compensating for the signal problems.
Even when you can see a mountain multiple times at once, the apparent distance to the mountain is going to have almost no variance as it shimmers around.
I understand that I am some random guy on the internet, but I've also spent a lot of time in Afghanistan (weird coincidence here, believe me or don't believe me) implementing two way satellite systems, and still work professionally in expensive two way satellite systems. I also happen to be a person who's been doing network engineering for ISPs since 1996-1997 or thereabouts, and currently has business cards that read "Senior Network Engineer" on them. I'm not willing to give up anonymity here, but ask a specific two way satellite technology question and I'll be more than happy to answer.
As a note there were lots of people buying and installing cheap COTS 1.2-1.8 meter size ku band vsat terminals at all sorts of military sites in Afghanistan for MWR purposes, it's not rocket science or really hard to do at all. This stuff was, and is, all kept air gapped and separated from anything that is a real DoD or ISAF/RS network. These were networks for peoples' own personal recreational internet and entertainment in their rest time. You absolutely would see weird latency, jitter, packet loss behavior on those.
Saying “you’re wrong” is not the same as saying “you don’t know anything about this topic”. Grandparent did the former; you’re accusing them of doing the latter.
Everybody can be wrong about something, no matter how smart / credentialed / experienced they are.
> I like that some random guy on the internet is telling a guy who was in charge of the satellite internet on a military base that he doesn't know anything about satellite internet.
I like that to you only one is a random guy on the internet. And I should know, I'm the president of the World Liking Things club, and you random guy on the internet have no business liking things like I do.
The military uses a hell of a lot of iridium, but from context it really sounds to me like he's describing the low end of the COTS C/ku band ISP market.
Until very recently you have not been able to move more than 2400 bps through an individual iridium link, limiting it to very low data rate IP networking and single user at a time per device.
The military itself uses a lot more geostationary than you might think at first glance. They don't have their own iridium sized low earth orbit narrow band networks. There's many series of DoD satellites in geostationary, and they also lease a lot of commercial transponder capacity.
We definitely lost both of our internet uplinks at our California farm when the wildfires filled the air with smoke this summer. One uplink is satellite and for weeks it had no signal as the farm is only a few kilometers from one of the largest fires.
The other uplink was a solar powered 4G connection and despite several deep cycle batteries and a few 200 watt panels, the station did not get enough light to stay powered over those weeks.
We finally found a hilltop wireless ISP that can serve us. We’ll try to get Starlink too but I don’t expect that to work well during fire season. It’s tough to be offline for weeks at a time.
I found a youtuber that has Starlink and had them run the dslreports bufferbloat test. They received an "A". This suggests that the jitter should be acceptable to most people on Starlink. Note: If running the test on Starlink, select "cable modem" as the Satellite tests on dslreports have not been updated to deal with Starlinks different behavior.
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.
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.
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.
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...
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.
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.
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.
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.
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.
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.
Is anyone else amazed that we are sending data to a satellite orbiting in space and getting a response that is 50-100 times faster (down) than the T1 links (1.544 mbps) our schools paid crazy amounts for when I was in high school almost 20 years ago? I mean that's pretty cool when you look at what Space X has done.
We certainly should be but I think we’re so used to the crazy advances over the last 30 years these things become mundane. I mean is it really not just as amazing that we have neural nets running on machines in our pockets when these were considered computationally too expensive for doing basic handwriting processing on super computers a couple decades ago?
how many satellites are up now & how many users are there?
what happens to get to the next 10x users? 100x?
it's cool that latency seems good.b that is the very first hurdle to viabity. but what hope there is for the future for this being real, useful, affordable broadband I think is still completely unknown.
whatever is happening today is being done at a radical loss, with a radically undersubscribed batch of satellites that is in no way going to reflect what kind of numbers people see if this goes anywhere, succeeds. maybe it can keep acting growing accommodating, providing magnitudes more bandwidth to people. maybe. while retaining this so far moderate price. maybe it's possible this works. but how it grows to serve orders of magnitude more folk, well, there are some constellation size asks, some guidelines, but it's very much indeterminate that the price point & offering now is viable, is really for real. there's a lot more finding out to do.
I agree. This to me is the major problem with starlink. Each sat only has 10-20gigabit/sec capacity, which really isnt a lot considering Netflix can use 15mbit/sec per stream, and call of duty patches can be nearly 200GB. It's not going to take many users to completely saturate 20gbit. It's not far off what a single 4G+5G cell site has, capacity wise, but instead of serving a few km max we are talking 100s of km^2.
I think it's highly likely there will be data caps and/or fairly serious traffic shaping on the production version at some point.
The subscription ratio (people: satellite) is a big one.
I'm also quite nervous about uplinks. According to dlgeek there are 893 satellites up right now [1]. If we grow to 10x that number, and each satellite can do 20Gbit (where is that number from, btw?), how do we begin to provide that throughout, whatever fraction of that max 2.2Tbit/s aggregate downlink for the satellite fleet? What is it going to take to literally uplink such huge volumes? I'm not well versed in satcom & how much spectrum is available, & how good ground stations are at pointing. Can one base station potentially fully saturate 10 or 50 or 250 satellites, once the sky starts to fill? Maybe we are that good, maybe we can shoot tiny narrow high-bandwidth wide-spectrum needles exactly as needed up into the sky, en masse, dozens of high bandwidth beams emmentating from each ground station. Or maybe we just need millions of ground stations, but can work that.
Does even that work? I keep hearing of "constellations," trains of satellites. How close will satellites stick? Can a ground station reuse spectrum across multiple transmitters seen talking to constellation members? or are the satellites too close, lumped together? this filters back to a previous question, how much spectrum is available for uplinks?
Trying to imagine the scale of this project, seeing both how we get data to in to the sky, then down to folks, it's epic. All the "zero to one, one to many" modes of growth discussion feels short of the epic planning & demands it'd take to both hoist satellites up (which SpaceX Starlink seems to be doing!) & the supply network to make the satellites useful.
But what kind of limits does such an attempt run into? Can you just keep adding more and more dishes, as long as you please, each one pointing at one satellite, with zero interference from any other of your ground dishes? No, I rather suspect not. When you are uplinking, there's probably going to be other satellites also getting some of your signal, especially as we move from 1000, to 10,000, to beyond numbers of satellites. This is not my field, but I feel like there are some very interesting limits to uplinking, a lot of constraints, that the sky itself can only hold so much uplink capacity, and it'd be interesting to me at least to have some more back-of-napkin discussions on what those constraints really look like.
> When you are uplinking, there's probably going to be other satellites also getting some of your signal, especially as we move from 1000, to 10,000, to beyond numbers of satellites.
Only starlink is going to be using this frequency band, so it doesn't matter if it hits anything else.
And of the starlink sats, you're not going to have more than a handful in view at once, so aiming is easy. And as far as base stations go, you just need one every X hundred miles, with several dishes.
The problems you're imagining could be real on a sufficiently dense constellation, but there is no reason to go anywhere near that density, so it's not a problem anyone has to worry about.
Every X hundred miles feels kind of like where I was going. Even then, there's a very active question of how many uplinks each given uplink ("gateway") would need to make, to provide the massive amounts of bandwidth required. Anyhow, I found a couple numbers on spectrum that are semi-interesting to me to assess what spectrum resources Starlink has purchased. According to this site[1], Starlink has:
27.5 – 29.1 GHz and 29.5 – 30.0 GHz and 47.2 – 50.2 GHz and 50.4 – 51.4 GHz = 6.1GHz of spectrum for uplink from base-stations ("gateways") to satellites
10.7 – 12.7 GHz and 37.5 – 42.5 GHz = 7GHz of spectrum for downlink from satellite to user "terminals".
According to this page, a year ago they were authorized to run 6 gateways.
I'm not sure how much throughput they can push to a satellite with 6.1GHz of total spectrum. It's still interesting to me to try to consider how even the current 893 satellites get anywhere near enough throughput up to them to be well utilizing their own downlink capabilities. I'm still only inuiting, but it still feels like something doesn't add up, I still don't see how the fleet is going to get enough bandwidth. My previous image of a data-center bristling with dozens & dozens of dishes pointing & tracking upwards indeed doesn't seem quite right (and the side-lobe interference question I was asking about), but the alternative, of a massive network of ground stations each boasting colossal connectivity, seems almost equally hard to imagine.
> I'm not sure how much throughput they can push to a satellite with 6.1GHz of total spectrum.
The satellites are very far apart, so they can use the full 6.1GHz per satellite with near zero interference. At most you'll have two in the same part of the sky.
The earth's surface area is about 200 million square miles. If you have 20 thousand satellites, that gives you ballpark 100x100 miles per satellite. Roughly 100 miles of separation, 350 miles of height, very easy to isolate just one.
With this many satellites, if you had a ground station every 300 miles, it would have to connect to about 10 satellites. It would need 200Gbps of fiber, which isn't hard to do. You'd be using less than 25 terabits per second across the entire continental US. That's less than a single undersea cable. No colossal bandwidth necessary, even at unreasonable numbers of satellites.
https://www.ispreview.co.uk/index.php/2020/09/analyst-probes... has the 20gbps (I've seen it in other places too). I actually think it could be considerably less than that in reality, as that's probably the raw link speed. I don't know the modulation techinque that starlink uses, but I imagine some level of Forward error correction (5/6 seems common in other sat systems I've read about), which drops throughput by ~20%, plus other modulation overheads. I wouldn't be surprised if the actual real world speed per sat is half the max, possibly a lot worse in bad weather.
I don't think uplink is a problem. Even 2TBit/sec of internet capacity is cheap (especially compared to satelite launches) so I imagine this is a rounding error in the costs. I'm sure they have thought about the uplink spectrum requirements.
The above article mentions they think that starlink could serve 1.5million US customers when the full (12k) constellation is deployed.
Did they tell what gonna happen with geofencing after the beta?
I know a few people who would probably want that thing installed on their boats. For marine applications, it needs to work globally without any fences. Geofencing-wise, international waters == nowhere. In territorial or internal water you normally have some 4G from the land, not much need for satellites there.
The current generation of satellites don’t have laser interlink connections.
That means the connection needs to go directly from the ground, up to a sat, then immediately back to the SpaceX ground station. That means for a while you’ll need to be within a certain range of a ground station, so international waters won’t work for a while.
The marine “anywhere on earth application” won’t be really applicable until they get the next generation of satellites with laser interlinks launched.
Still, territorial waters extend for 12 nautical miles. Apparently, these satellites fly at 1100-1300 km. If ground stations (both SpaceX backbone routers, and mobile terminals) can talk to a satellite 45 degree above horizon, the satellite can connect the terminal with the ground station in a circle 1100-1300 km radius underneath (slightly less due to Earth curvature, but not by much).
Not gonna work in the middle of the Pacific, but might be good enough for many. For instance, one can sail from here across Mediterranean Sea all the way to Africa or middle East, while staying within reach of Europe-based routers. That’s assuming SpaceX gonna build them.
Some of them will fly even lower in the later phases.
But 45° is also wrong, they go as low as 25° above the horizon now. Which means that the maximum radius that a satellite can cover should be around 500km if my last calculations were correct and I remember correctly.
Under ideal conditions you could have internet up to 1000km off the coast then even without laserlink. Under normal conditions it would be more like 500km from the nearest ground station.
I'm still hoping that they can use user terminals as ground stations for these maritime scenarios.
The bent pipe nature is why globalstar was a dismal failure in the aviation and maritime markets, while the second corporate incarnation of iridium succeeded.
I don't doubt that SpaceX will figure out satellite to satellite high capacity trunk links within the next 5-6 years. But for now they are definitely proceeding with building medium sized starlink earth stations in the same general regions as where the CPE will be located.
Someone asked Elon on Twitter if they can use Starlink on their yacht while crossing the Pacific and Elon said that is a great use-case because there won't be many other connections to the same sat.
He has also said a few times it will work perfectly well on things like moving trains.
So I have to think the geofence will go away, though I assume there will be country restrictions due to legalities.
What seems to be missing from these discussions is the gigantic bet that SpaceX is making on Starlink. They performed 15 launches for this constellation already that they paid for themselves. If the internet service business doesn't generate revenue the launcher side will see a majority of its launch manifest evaporate and book enormous loss.
From the outside it seems that a commercial failure of Starlink would bankrupt all of SpaceX.
I think this is admirable. He's clearly very passionate about his companies and though I'm sure money is a motivating factor it's not his priority - he's said as much before and he doesn't exactly live a lavish lifestyle for a billionaire
Is there any elasticity to that demand, i.e., has SpaceX priced launches to only attract 12 so far this year, or would they still only attract that many if they lowered the price?
There is almost no elasticity in current market: DOD, NASA and traditional GTO comsat launches are few in number and not particularly sensitive to price. Payloads routinely cost several times the price of a launch so cheaper launches have little impact.
The satellite service business is much larger than the launch business and this is why SpaceX is attempting to enter.
> Payloads routinely cost several times the price of a launch so cheaper launches have little impact.
Little immediate impact. Of course, if launch gets way cheaper, the things you do to maximize longevity, fault-tolerance, and mass efficiency start to matter less...
That is, payloads are expensive in part because launch is expensive.
The problem is that the main business of SpaceX and Arianespace was GEO sats. Those have been in a slump and fewer launches existed. Maybe that will pick back up, but the largest part of that were for Internet and after that TV, both don't seem like huge growth markets.
Small sats have become so powerful, that on a SpaceX Rideshare you can deploy basically a whole constellation in one shot potentially. Most sats today are no longer that large and heavy that you have dedicated launches.
Many of their launches are government, both domestic and foreign.
SpaceX has not reduced the price 'that much' in the last couple of years. They were already the cheapest on the market and opted for more profit, so they could have better cashflow when going into Starship and Starlink.
What also has to be considered is that even that 12 is a pretty large number. Large rockets usually didn't fly that at the rate SpaceX flies them.
The sat market is slow the respond, 4 years are very little in it. Hopefully people will come up with more great ideas how to use the capability SpaceX offers.
OK, yes. Two of the launches had 58/60 Starlink and another very small satellite, and one had 57/60, so there were 3 launches that were only each 95% of a full Starlink launch. :P
It's likely the cost to launch is significantly lower, especially as they are often the 3rd/4th/5th/6th (has there been a 6th yet?) launch of a booster.
Assuming the $62m pricetag covers the cost of the booster, the marginal cost of the Starlink launches is recovery of booster and fairings + refurbishment of booster and fairings + stage 2 + fuel + operations.
A smallsat rideshare may well be paying a significant portion of those costs.
Yes, SpaceX probably has OK margins, but when we're considering how much of SpaceX's manifest is Starlink, this is not really relevant. There's been 9 launches + 3 launches that are 5% not Starlink = 9.15 launches worth of non-Starlink revenue.
And even having this different argument you're triyng to have, SpaceX's margins are not so good that collecting 5% of nominal market price will put them in the black for that launch... That's just fantasy.
I generally agree with your points, and reading back realised I misread you; you said 3-7% of the rate for an entire launch, and I read that as 3-7% of that launch.
On the broader point, I agree that SpaceX is sinking a lot into the Starlink, but even if it fails I doubt bankrupts the company. Part of that is just how cheap launch is for them, but it’s also because I can’t see a catastrophic fail mode for them. The satellites are clearly already working, worst case they have to sell the constellation to someone else. Probably a failure of imagination on my part though.
I believe the business case for Starlink is solid, my parents currently live in a rural area with dial up. They have a few dozen neighbors all on dial up.
They currently pay $70 a month for their phone lines, and I bet a couple million Americans are in the same boat. They could probably cancel their entire bill with ATT if Starlink is reliable, as digital phone service is extremely cheap.
They're also about to tap into billions in FCC rural subsidies. If they can get subsidized pricing down to, say, $100 for equipment and $50/month they will have a lot of demand.
The dish has a phased array antena already, so it is already tracking the satellites. Likely such approach, unless integrated in a bigger dish of some kind to increase signal to noise ratio, already is much more precise and faster than any physical moving mechanism.
159 comments
[ 2.7 ms ] story [ 175 ms ] threadIsn't that true of every product?
It's still very easy to forget.
For those with marginal service, this is a material improvement. It’s also firming up the case that StarLink will generate billions of dollars per year in revenue (eclipsing SpaceX’s launch revenues), from both government contracts (rural broadband and DoD) and unsubsidized subscribers. This doesn’t include revenue from StarLink’s contract building missile tracking satellites for the Defense dept (based on StarLink sats) [2].
[1] https://arstechnica.com/tech-policy/2020/10/spacex-gets-fcc-...
[2] https://www.theverge.com/2020/10/6/21503165/spacex-space-dev...
It's even better than DSL and some wireless ISPs or 4G hotspots with marginal connections, which are probably the main competition for Starlink.
The whole reason CowboyNeal was attacking the iPod was that technically it was a step backwards. It couldn't do anything that a Nomad couldn't. The part that it did well (and that he missed the importance of) was the usability and polish.
> Compared to existing satellite ISPs you get dramatically better latency and dramatically faster/easier setup (no professional installation required, plug and play in 5 minutes). It's also likely to be far cheaper if you actually transfer a decent amount of data, and early tests seem to show better reliability in poor weather as well.
All true, but can you see HN caring about any of that if SpaceX wasn't involved? The HN userbase is overwhelmingly city folk; I suspect only a tiny proportion of us have ever used satellite internet before.
these satellites are radically undersubscribed right now, serving a handful of users tops. that's not going to model what happens if this works & grows.
The source used for the comment "every 5-10 minutes" doesn't fully clarify that point. Many users in the Reddit subreddit post that is quoted stated that the person likely installed it incorrectly and the cutouts were from trees blocking the signal. So "every 5-10 minutes" is not what most people are seeing. Source reddit post the article uses: https://www.reddit.com/r/Starlink/comments/jsnd52/starlink_d...
Key reply:
> Based on the disconnects every 5-10 minutes. I would definitely say the obstructions are causing issues. I had the same problem last week and timing between disconnects. Put mine a little too close to the house. Moved it yesterday and today my connections were holding with no disconnects for anywhere from 2 to 3 hours with maybe a short 10-30 second disconnect between.
Sure, in theory, the entire world can have access but that won't come for free and won't be in any way affordable by the vast majority of the world population.
https://www.pcmag.com/news/native-american-tribe-gets-early-...
Satellites are small... and 2 miles up, the world is even bigger than it is down here.
I think you have a pretty misguided view of the developing world and their buying power. There are a lot of middle-income countries with reasonable spending power, but lousy infrastructure, in-between the extremes of the undeveloped Congo (which yes, will take some time to get Starlink penetration) and residents of a NYC condo.
Another thing you ignore is that if you combine Starlink with a mobile station you can potentially provide internet to a whole village.
Also, the prices will likely not be the same everywhere. SpaceX will want to get higher utility of the overall system and having unused sats fly over Africa is pointless. Lowering the price in some regions to get full utilization if an economic non-brainier if they would actually have issue selling the service.
In addition, the idea that this service wouldn't be worth it for rural people in the first world is very questionable. Don't rural people deserve good internet?
Eg. It costs $50 per month for regular speed, $100 for double regular speed, etc.
You will hopefully get a better signal by having many sats in your view and the ability for the network to make smart choices about where to send what when.
For the beta it seems to be tied to a geographic location.
(other than the US, if it so chooses)
Legal issues aside, they're also quite literally beyond the reach of all but a small handful of spacefaring nations, and even a nation like China that has the technical capability would not resort to actively shooting them down unless other avenues were exhausted.
- Ground station
- SpaceX allowing you access from that location
Some beta testers have moved around their antennas and they worked even when not at the 'official' location. So currently it 'seem' like they are just relaying on beta tester to be mostly honest.
Once you have real sat-to-sat laser comms (next generation of sats, currently being tested) you can use it anywhere theoretically.
We don't know what the long term strategy is around geo-blocking. In theory they could have an issue where some location is fully subscribed and then somebody is moving there causing an issue. It will be interesting to see what policy they will use.
Source: I ran a small satellite ISP on a base in Afghanistan.
But he wrote that he'd used "TCP over UDP". The first lines of the TCP over UDP spec already state
"It provides exactly the same congestion control, flow control, reliability, and extension mechanisms as offered by TCP." [1]
[1] https://tools.ietf.org/id/draft-baset-tsvwg-tcp-over-udp-01....
TCP is designed for encapsulation in IP, this is why it is called TCP/IP. And UDP is very similar to IP, only difference is that UDP gives you source and destination port numbers and a payload checksum, where IP only gives you a protocol numbers and non-checksummed data. None of these "features" is going to have any effect on how TCP is delivered.
So the only effect that use of TCP over UDP would have here is:
- increase the overhead used for headers, thereby decreasing the MSS (segment size)
- "cloak" the TCP traffic as UDP and thereby prevent any routers and/or middleboxes [2] from applying special treatment to TCP (i.e. traffic shaping, buffering etc.)
Maybe the latter effect is what made a difference for the link in question.
[1] https://tools.ietf.org/id/draft-baset-tsvwg-tcp-over-udp-01....
[2] https://en.wikipedia.org/wiki/Middlebox
edit: formatting
When dealing directly with UDP, one can implement different application-level logic for dealing with packet loss: "NAK" packets, different backoff strategies, and even forward error correction.
If you saw a lot of jitter it was because you were in an oversubscribed, contended tdma network. What was your terminal hardware, what modems were you using and what was the rf chain and modem setup at the far end (not in Afghanistan)?
Latency on terrestrial ptp microwave also does not vary. Ducting and diffraction and temperature inversions on long 6GHz links affects rsl and modulation, not latency.
In the case of starlink, each satellite is functioning as a bent pipe relay between cpe and starlink earth station. Latency will vary a ms or two based on satellite position during a pass overhead, but no more than that. The rest of the latency is the same as from any terrestrial internet customer in the same geographic location as the earth station. For instance one of the beta test earth stations is in north bend, wa, colocated with a level3 long haul dwdm site.
People in WA have been posting tests with latencies of 25-26ms to speed test servers in downtown Seattle. That's actually really, really good, considering that latency to your default gateway on a residential cable docsis3 coaxial connection can easily be 16ms on a terrestrial connection. North bend is a few ms from seattle, the brewster starlink teleport site several ms more.
In the case of starlink they are competing against 32:1 or worse oversubscribed geostationary consumer grade vsat services like viasat and Hughesnet. Starlink is far superior tech.
Edit: And, any really good blog series on this? so I know what to expect in such a course
But my knowledge of networks is fairly rudimentary and old now and now that I think about it, pings would get sudden spikes in the rain too and that's just a lower layer than I'm talking about. hmm. I was also a child then so I might be misremembering.
For an example of a modern radio google aviat wtm4200.
[1] https://operator.ui.com/ptp-bridging
For something like the af5xhd, it's a great little thing for its price, and with good quality antennas at both ends can make a very effective bridge. But for more serious things? No. It's so bad they haven't even been able to implement snmpv2 on their $1000+ radios. The af11 is so bad it only uses a 56MHz channel mask in an 80MHz channel.
No matter what manufacturer, serious ptp radios can be distinguished by two things: the presence of terminals for a direct -48vdc wiring connection and sfp/sfp+ ports.
At the very low end of cost, ubnt's 802.11ac chipset ptp bridge radios are great and very cost effective, I'm talking about the powerbeam ac series, nanobeam and similar. For when you want to buy a few hundred bucks of stuff on your credit card to link a few buildings together quickly. For anyone looking at that category of stuff they should also look at mikrotik and mimosa radios.
Lost packets increase latency. And weather conditions cause lost packets.
Only on HN
That said unless we're talking about lasers air refraction doesn't usually impact RF(there is tropospheric ducting but that's something different).
Latency as experienced by users might be much worse, however, if suddenly 300Mbps of traffic is trying to be shoved through a 40-75Mbps link, for a period of several minutes or more.
From the point of view of an end user on a LAN at a site dependent on two way satellite it may not be possible for them to determine exact why they're seeing packet loss.
Ultimately, the latter is the kind of latency the user cares about.
They can both be right as walrus01 seems to talk about latency of a single rtt in isolation, and withinboredom about what does the TCP endpoint effectively see.
Example: at one location, a 3.8 meter ku band dual linear polarity compact cassegrain dish in somalia, a 200W BUC+SSPA, good quality norsat PLL LNB, a cdm760 modem. Ten MHz of transponder space occupied total. Other end, a 5.5m dish with similar gear on it in a teleport in Cyprus.
If the person I'm replying to isn't willing or able to clarify whether they were in a contended, or dedicated network, or specify the gear, it's hard to say exactly why they're seeing jitter. But I can quite confidently say that excursions above 495ms are not something seen in a dedicated network.
It's how that dedicated capacity can get sliced up and shared that is the most common cause of packet loss, jitter and high latency in geostationary based two way satellite IP links.
The most common thing for a cheap VSAT isp to do is to acquire one chunk of Ku transponder capacity (let's say about $8000 a month worth), install something like an idirect hub with line cards in it, and sell access to individual end users in $developingnation at an 8:1 oversubscription ratio or more. Meaning that an end user vsat terminal might see 495ms at a lucky time when its peers in the same network were not also moving traffic, but could also often see latency anywhere from 500 to 850 during busy periods.
Now that same 8000 bucks a month of transponder space could equally well be assigned to a single high priority end user with money to pay, in which case they would never see above 495ms. Or it could be oversubscribed to all hell, such as if the ISP wanted to sell $280/month service to a bunch of very cost conscious end users in a developing nation environment who can only afford very cheap VSAT terminal hardware (one 1.2m fiberglass compression molded reflector, a 4W BUC, and an idirect modem) and minimal monthly ongoing payments. In which case each user will be in a 16:1 or worse oversubscribed network.
The economics of making a 5500 kilogram $200 million satellite in geostationary orbit pay for itself, before its lifespan is done in 15 years, are not favorable. If you ever want to truly be shocked go price the $ per Mbps per month that results from the 1:1 dedicated setup I described above.
On the topic of what the average hn reader considers "satellite internet". Based on consumer experience with the absolute cheapest consumer grade ku and Ka band vsat terminals in the usa, they almost 99% of the time mean Hughesnet or viasat/wildblue. Very few people actually have money or reason to lay their hands on real two way satellite stuff. Generally if you work for a company that needs something like a $3500/month link for a mine in Nunavut you're paying some third party experts to implement it.
Note that may be true for one bird in Afghanistan, but the path length isn't exactly the same everywhere, nor are the delays of all modulaiton schemes.
For instance our site in Germany was more than 15ms away by fiber from DE-CIX stuff in Frankfurt. The site in Singapore had better latency to any cdn or hosting operation that was in Singapore itself, but much worse if traffic had to find its way through the terrestrial submarine fiber internet to other places. End to end latency between Afghanistan-germany-seattle or Afghanistan-singapore-seattle was considerably different.
Yes, the position of the two end point sites on the sphere that is the earth will have slightly different speed of light times to a satellite in geostationary orbit. But 99% of that latency in the above 490ms figure is the extreme distance of the satellite, and not whether or not one site is slightly further down the sphere, as viewed from a theoretical observer sitting on the satellite.
It is also worth noting that modern scpc/mcpc modems can be intentionally configured to use more computationally "expensive" FECs, in a trade off of slightly more bps/Hz squeezed into a given sized slice of transport kHz, at the expense of a few ms higher latency. This is a network engineering design decision made when a link is first brought online. Modern modems and ASICs, FPGAs have made this less difficult than it used to be. If I can use an analogy for software developers, this is sort of like the speed vs file size trade off choice made between using gzip, bzip2, xzip, at different settings and compression levels.
Path length varies from ~34500km to ~42000km (~115ms to ~140ms). So, assuming an ideally placed teleport, from 115 * 4 = 460ms to 115 * 2 + 140 * 2 = 510ms speed of light delay on round trips. This is why I said I found 492ms to be an amazingly over-precise number. Just call it 500ms. :P
> 3.8 meter ku band dual linear polarity compact cassegrain dish > a 5.5m dish
Are these receiver sizes in the range of what Starlink expects from users? I had in mind something more like a DirectTV dish. If the consumer dish sizes are indeed smaller, how does that affect the discussion about jitter and other aspects of link quality?
The starlink cpe is indeed in the category of directv dish sized.
With only a tiny bit of the relevant subjects under my belt, I would expect the metrics to be significantly worse using a much smaller dish. But it seems like you're suggesting Starlink could achieve similar metrics on a smaller dish? I'm curious how that works. Apologies if I'm misunderstanding.
Let me see if I can answer some of this.
We were on a 1:1, eventually. Tried the 8:1 (or something like that) at first and it was terrible.
Latency was almost always, like you said, almost exactly 495ms at the link. However, we'd suffer transient jitter that directly correlated to weather in the atmosphere. Essentially, satellite communication is a laser (parabolas ftw), so this made sense at the time. Correlation doesn't necessarily mean causation, but it was annoyingly correlated.
I only made this correlation because my "day job" in the military had a lot to do with radar, microwave antennas, and optics. So, I'd observe the refraction personally (being able to see mountains as clutter for example, when the radar rays were pointed well over the mountains) and then notice the observed jitter on the network. Not to mention people coming to tell me how unusable Skype was at the moment and could I fix it please?
No offense, but you probably don't see this at your level as a network engineer unless you're "boots on the ground." When you're there and easily accessible, people will come complain to you when any and every little thing that goes wrong and "their wife is threatening to leave them and they can't even have a conversation, omg, fix it right now plz."
In a more corporate setting, people are far less likely to complain about transient issues because they fix themselves before the complaint can be made, or if they do, it's easy to dismiss as other issues.
To be honest, we had those other issues as well. QoS was a thing, and constantly battling bittorrent clients and their users was also a thing. But those issues were usually only during the evenings when lots of people were trying to call home and get on Facebook. I'm talking about when none of those issues were at play, such as late at night or during the day.
The UDP solution was directly related to periods of saturation, specifically I built a custom transparent proxy forked from Squid. It's sole job was to act as a mediator between a server back in the US that did a few things:
1. On the Afghanistan side, it would wrap the HTTP(S) request packet and send it via a TCP control channel to the server in the US.
2. On the US side, it would do some image compression (if applicable) and other things to the response to make it smaller (in bytes).
3. On the US side, it would also keep track of user flows and how much bandwidth each user got, and spit back UDP packets, reserving a dynamic % of bandwidth for control and unregulated packets. There was also an open TCP control connection for sending back smaller things that fit within a few packets.
4. On the Afghanistan side, it would put all these packets together to reform the stream, keeping track of any dropped packets (rare) and then forwarding the completed stream (or once the buffer was full) to the user at 100mbps. If any were dropped, it sent back a bitmap over the open TCP connection to have those frames retransmitted via TCP and then inserted into the stream before going to the client.
It worked fantastic for things like youtube, since the subscriber could make full use of their subscribed bandwidth with a half-second delay after clicking a link. The only things that didn't use that system were things like Skype. We never really had transport-related congestion issues after this was built as a hack, over several weeks. To be honest, it probably would have worked just fine by forwarding over already open TCP connections to the back-end server and giving each subscriber a dedicated stream. However, time in the desert is a premium and once something works good-enough, it pretty much stays that way forever.
> If the person I'm replying to isn't willing or able to clarify whether they were in a contended, or dedicated network, or specify the gear, it'...
Even when you can see a mountain multiple times at once, the apparent distance to the mountain is going to have almost no variance as it shimmers around.
As a note there were lots of people buying and installing cheap COTS 1.2-1.8 meter size ku band vsat terminals at all sorts of military sites in Afghanistan for MWR purposes, it's not rocket science or really hard to do at all. This stuff was, and is, all kept air gapped and separated from anything that is a real DoD or ISAF/RS network. These were networks for peoples' own personal recreational internet and entertainment in their rest time. You absolutely would see weird latency, jitter, packet loss behavior on those.
Everybody can be wrong about something, no matter how smart / credentialed / experienced they are.
I like that to you only one is a random guy on the internet. And I should know, I'm the president of the World Liking Things club, and you random guy on the internet have no business liking things like I do.
Until very recently you have not been able to move more than 2400 bps through an individual iridium link, limiting it to very low data rate IP networking and single user at a time per device.
The military itself uses a lot more geostationary than you might think at first glance. They don't have their own iridium sized low earth orbit narrow band networks. There's many series of DoD satellites in geostationary, and they also lease a lot of commercial transponder capacity.
The other uplink was a solar powered 4G connection and despite several deep cycle batteries and a few 200 watt panels, the station did not get enough light to stay powered over those weeks.
We finally found a hilltop wireless ISP that can serve us. We’ll try to get Starlink too but I don’t expect that to work well during fire season. It’s tough to be offline for weeks at a time.
Irrespective of how right or wrong your first paragraph is; this is one of the reasons I absolutely love the HN community.
[1] - http://www.dslreports.com/speedtest
No doubt the scummy high frequency traders are already jockeying to get out in front of this change in the telecom industry.
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 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.
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.
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.
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.
what happens to get to the next 10x users? 100x?
it's cool that latency seems good.b that is the very first hurdle to viabity. but what hope there is for the future for this being real, useful, affordable broadband I think is still completely unknown.
whatever is happening today is being done at a radical loss, with a radically undersubscribed batch of satellites that is in no way going to reflect what kind of numbers people see if this goes anywhere, succeeds. maybe it can keep acting growing accommodating, providing magnitudes more bandwidth to people. maybe. while retaining this so far moderate price. maybe it's possible this works. but how it grows to serve orders of magnitude more folk, well, there are some constellation size asks, some guidelines, but it's very much indeterminate that the price point & offering now is viable, is really for real. there's a lot more finding out to do.
I think it's highly likely there will be data caps and/or fairly serious traffic shaping on the production version at some point.
I'm also quite nervous about uplinks. According to dlgeek there are 893 satellites up right now [1]. If we grow to 10x that number, and each satellite can do 20Gbit (where is that number from, btw?), how do we begin to provide that throughout, whatever fraction of that max 2.2Tbit/s aggregate downlink for the satellite fleet? What is it going to take to literally uplink such huge volumes? I'm not well versed in satcom & how much spectrum is available, & how good ground stations are at pointing. Can one base station potentially fully saturate 10 or 50 or 250 satellites, once the sky starts to fill? Maybe we are that good, maybe we can shoot tiny narrow high-bandwidth wide-spectrum needles exactly as needed up into the sky, en masse, dozens of high bandwidth beams emmentating from each ground station. Or maybe we just need millions of ground stations, but can work that.
Does even that work? I keep hearing of "constellations," trains of satellites. How close will satellites stick? Can a ground station reuse spectrum across multiple transmitters seen talking to constellation members? or are the satellites too close, lumped together? this filters back to a previous question, how much spectrum is available for uplinks?
Trying to imagine the scale of this project, seeing both how we get data to in to the sky, then down to folks, it's epic. All the "zero to one, one to many" modes of growth discussion feels short of the epic planning & demands it'd take to both hoist satellites up (which SpaceX Starlink seems to be doing!) & the supply network to make the satellites useful.
[1] https://news.ycombinator.com/item?id=25109020
Put a lot of sat dishes on the top of major Internet exchanges such as DE-CIX and on all data centers of AWS, GCE and Azure.
But what kind of limits does such an attempt run into? Can you just keep adding more and more dishes, as long as you please, each one pointing at one satellite, with zero interference from any other of your ground dishes? No, I rather suspect not. When you are uplinking, there's probably going to be other satellites also getting some of your signal, especially as we move from 1000, to 10,000, to beyond numbers of satellites. This is not my field, but I feel like there are some very interesting limits to uplinking, a lot of constraints, that the sky itself can only hold so much uplink capacity, and it'd be interesting to me at least to have some more back-of-napkin discussions on what those constraints really look like.
Only starlink is going to be using this frequency band, so it doesn't matter if it hits anything else.
And of the starlink sats, you're not going to have more than a handful in view at once, so aiming is easy. And as far as base stations go, you just need one every X hundred miles, with several dishes.
The problems you're imagining could be real on a sufficiently dense constellation, but there is no reason to go anywhere near that density, so it's not a problem anyone has to worry about.
27.5 – 29.1 GHz and 29.5 – 30.0 GHz and 47.2 – 50.2 GHz and 50.4 – 51.4 GHz = 6.1GHz of spectrum for uplink from base-stations ("gateways") to satellites
10.7 – 12.7 GHz and 37.5 – 42.5 GHz = 7GHz of spectrum for downlink from satellite to user "terminals".
According to this page, a year ago they were authorized to run 6 gateways.
I'm not sure how much throughput they can push to a satellite with 6.1GHz of total spectrum. It's still interesting to me to try to consider how even the current 893 satellites get anywhere near enough throughput up to them to be well utilizing their own downlink capabilities. I'm still only inuiting, but it still feels like something doesn't add up, I still don't see how the fleet is going to get enough bandwidth. My previous image of a data-center bristling with dozens & dozens of dishes pointing & tracking upwards indeed doesn't seem quite right (and the side-lobe interference question I was asking about), but the alternative, of a massive network of ground stations each boasting colossal connectivity, seems almost equally hard to imagine.
[1] https://www.elonx.net/starlink-compendium/
The satellites are very far apart, so they can use the full 6.1GHz per satellite with near zero interference. At most you'll have two in the same part of the sky.
The earth's surface area is about 200 million square miles. If you have 20 thousand satellites, that gives you ballpark 100x100 miles per satellite. Roughly 100 miles of separation, 350 miles of height, very easy to isolate just one.
With this many satellites, if you had a ground station every 300 miles, it would have to connect to about 10 satellites. It would need 200Gbps of fiber, which isn't hard to do. You'd be using less than 25 terabits per second across the entire continental US. That's less than a single undersea cable. No colossal bandwidth necessary, even at unreasonable numbers of satellites.
I don't think uplink is a problem. Even 2TBit/sec of internet capacity is cheap (especially compared to satelite launches) so I imagine this is a rounding error in the costs. I'm sure they have thought about the uplink spectrum requirements.
The above article mentions they think that starlink could serve 1.5million US customers when the full (12k) constellation is deployed.
I know a few people who would probably want that thing installed on their boats. For marine applications, it needs to work globally without any fences. Geofencing-wise, international waters == nowhere. In territorial or internal water you normally have some 4G from the land, not much need for satellites there.
That means the connection needs to go directly from the ground, up to a sat, then immediately back to the SpaceX ground station. That means for a while you’ll need to be within a certain range of a ground station, so international waters won’t work for a while.
The marine “anywhere on earth application” won’t be really applicable until they get the next generation of satellites with laser interlinks launched.
Still, territorial waters extend for 12 nautical miles. Apparently, these satellites fly at 1100-1300 km. If ground stations (both SpaceX backbone routers, and mobile terminals) can talk to a satellite 45 degree above horizon, the satellite can connect the terminal with the ground station in a circle 1100-1300 km radius underneath (slightly less due to Earth curvature, but not by much).
Not gonna work in the middle of the Pacific, but might be good enough for many. For instance, one can sail from here across Mediterranean Sea all the way to Africa or middle East, while staying within reach of Europe-based routers. That’s assuming SpaceX gonna build them.
SpaceX updated all of their applications to only request 550 km.
Under ideal conditions you could have internet up to 1000km off the coast then even without laserlink. Under normal conditions it would be more like 500km from the nearest ground station.
I'm still hoping that they can use user terminals as ground stations for these maritime scenarios.
I don't doubt that SpaceX will figure out satellite to satellite high capacity trunk links within the next 5-6 years. But for now they are definitely proceeding with building medium sized starlink earth stations in the same general regions as where the CPE will be located.
He has also said a few times it will work perfectly well on things like moving trains.
So I have to think the geofence will go away, though I assume there will be country restrictions due to legalities.
On a personal boat this is less of an issue, but for most large boats the operate would install it and then offer local WLAN I would expect.
From the outside it seems that a commercial failure of Starlink would bankrupt all of SpaceX.
He has said many times the slower-than-ideal ramp for the Model 3 assembly line very nearly bankrupted Tesla (it was within a few months).
When the first two SpaceX launches failed he spent all their remaining money on the next one - if it failed, SpaceX would have too.
He has also said if either Tesla or SpaceX go bankrupt, so does he.
He is literally betting the house with every major decision.
Can you clarify this? I don't know much about it but can't see how they relate.
12 were Starlink.
About half of SpaceX's launch manifest is launching their own satellites for Starlink.
The satellite service business is much larger than the launch business and this is why SpaceX is attempting to enter.
Little immediate impact. Of course, if launch gets way cheaper, the things you do to maximize longevity, fault-tolerance, and mass efficiency start to matter less...
That is, payloads are expensive in part because launch is expensive.
Small sats have become so powerful, that on a SpaceX Rideshare you can deploy basically a whole constellation in one shot potentially. Most sats today are no longer that large and heavy that you have dedicated launches.
Many of their launches are government, both domestic and foreign.
SpaceX has not reduced the price 'that much' in the last couple of years. They were already the cheapest on the market and opted for more profit, so they could have better cashflow when going into Starship and Starlink.
What also has to be considered is that even that 12 is a pretty large number. Large rockets usually didn't fly that at the rate SpaceX flies them.
The sat market is slow the respond, 4 years are very little in it. Hopefully people will come up with more great ideas how to use the capability SpaceX offers.
Falcon 9 launch prices are about $62M.
So, Skysat probably paid 3-7% of the rate for an entire launch.
Indeed, see https://spacenews.com/spacex-rideshare-program-putting-downw...
It's likely the cost to launch is significantly lower, especially as they are often the 3rd/4th/5th/6th (has there been a 6th yet?) launch of a booster.
Assuming the $62m pricetag covers the cost of the booster, the marginal cost of the Starlink launches is recovery of booster and fairings + refurbishment of booster and fairings + stage 2 + fuel + operations.
A smallsat rideshare may well be paying a significant portion of those costs.
Yes, SpaceX probably has OK margins, but when we're considering how much of SpaceX's manifest is Starlink, this is not really relevant. There's been 9 launches + 3 launches that are 5% not Starlink = 9.15 launches worth of non-Starlink revenue.
And even having this different argument you're triyng to have, SpaceX's margins are not so good that collecting 5% of nominal market price will put them in the black for that launch... That's just fantasy.
On the broader point, I agree that SpaceX is sinking a lot into the Starlink, but even if it fails I doubt bankrupts the company. Part of that is just how cheap launch is for them, but it’s also because I can’t see a catastrophic fail mode for them. The satellites are clearly already working, worst case they have to sell the constellation to someone else. Probably a failure of imagination on my part though.
They currently pay $70 a month for their phone lines, and I bet a couple million Americans are in the same boat. They could probably cancel their entire bill with ATT if Starlink is reliable, as digital phone service is extremely cheap.
This cloud be a nice little DIY project.
This assumes that laser links are working which they are not currently.