4 ms·

I will confess that I don't get it and I would appreciate thoughts on why this is a good idea instead of a FOMO kind of thing (and especially AWS Kuiper), The

I will confess that I don't get it and I would appreciate thoughts on why this is a good idea instead of a FOMO kind of thing (and especially AWS Kuiper), The total switching capacity of the entire __planned__ constellation is around that in a _single_ large switch chassis or internet core router and the individual satellites have the capacity of a beefy enterprise Wifi6 access point; however the geographic area each will service is enormous. What I can't figure out is: what is the market? The market for an Iridium-style remote access is demonstrated to be tiny, there just is not the bandwidth or capacity to target "unserved" areas for ordinary internet. So what's left that's big? Industrial IoT? Military? Even industrial placements (Rio Tinto, etc. mining concerns) will prefer to set up dedicated links and regional wireless to this solution, so IIOT use case is really a pre-development stage market, which is again small. I even went looking for what the business press said (example: [1]). Reading this almost makes me think the author is in on a sort of dark joke; especially the “Fortunately, companies have new options for generating revenue from connectivity” section appears to be written by someone with a very dark sense of humor who is straight up trolling the reader. [1] https://www.mckinsey.com/industries/aerospace-and-defense/our-insights/large-leo-satellite-constellations-will-it-be-different-this-time https://www.mckinsey.com/industries/aerospace-and-defense/ou... edit: added "planned" to distinguish the long term from the immediate term [the first batch Starlink satellites didn't even do any in-space switching at all]..

jeffnappi 6y ago
Can you point me to your source on the planned constellation total switching bandwidth? I haven't read about that anywhere. I have read that they don't expect to be able to provide high levels of service in dense urban areas. Which does make me question the business model... I think you nailed it that the military and government will be a big customer. Iridium has ~1300ms+ latency and very low data rates, very high costs and the government is their largest customer from what I understand. Airlines are another likely market that would be a significant amount of business (post-covid19). If Starlink can compete against Iridium and the other existing satellite providers on price, they could at a minimum take most of that business.
- foobiekr 6y ago
  Per satellite max bandwitdh, per SpaceX, at commercialization will be a max of 17gbps [1, sorry for the lame source but it's what I could find at the moment]. Depending on how they're measuring that, the actual throughout could be substantially less than that. Given that their FCC filing was 20gbps per satellite, I think that's the raw, uncorrected bit rate, not the corrected bitrate which will be lower. A coworker and I did a back-of-envelope guess based on their power budget (switching capacity in bandwidth and pps and power are basically directly tied, ditto capacity and spectrum) and that's about what we arrived at as their upper bound (imho, not even half of that will be achievable). Note that this is a double-counted metric (bps total is bps_in+bps_out) for actual traffic volume. So let's go with that. The current SpaceX fantasy business model is to build out a constellation of 30,000 satellites, but near term not even close to that. Being very generous, let's credit them with an incredible 10k satellites, pretend they are not power constrained and use the full 20gbps allotted to them, and you get a total of 200k gbps total capacity for the constellation, uncorrected, assuming ample power, or on a good day 125Tbps for real data. Using a somewhat older core switch, the Juniper QFX10016, can do line rate switching for 480 100GE ports; 48Tbps in a single 21RU chassis from 2017. [2] These are commonly deployed in pops. The next generation will be doing that for 400GE ports in the same form factor, 192Tbps. A Cisco 8818, which is shipping right now, does 260Tbps per chassis in a 33RU form factor. The point is, anyone claiming we could serve even a large metro with a single one of these switches, let alone planet earth, would be laughed at, yet here we are. A given nation-scale network will have hundreds of routers with these kinds of capacities. I also don't think Starlink will get anywhere near the estimates I'm using here. [1] https://www.teslarati.com/spacex-starlink-satellite-upgrade-more-bandwidth-more-beams/ https://www.teslarati.com/spacex-starlink-satellite-upgrade-... [2] https://www.juniper.net/documentation/en_US/release-independent/junos/topics/topic-map/qfx10016-system-overview.html#power-supplies https://www.juniper.net/documentation/en_US/release-independ... [3] https://www.cisco.com/c/en/us/products/collateral/routers/8000-series-routers/datasheet-c78-742571.html https://www.cisco.com/c/en/us/products/collateral/routers/80... edit: moved the [2] to the right place. edit damage.
  - jeffnappi 6y ago
    Thanks for the data points :) I think something to consider here is that there is an iterative development process taking place with the satellites. Given the stated lifespan of the satellites is 3-4 years, they likely expect to see additional innovation over time and are gambling that the network will be improved over time, and may one day actually be capable of serving urban areas. This seems like a wise approach and appears to work well for both SpaceX and Tesla. The most viable use case in the near term is clearly to serve remote and rural areas which is admittedly a small market. There is a chance you might see governments subsidize service to rural areas that are underserved. I also imagine there will be a number of marine and aviation customers - though that's likely even smaller than the rural market. But yeah, I'm not one of the people who will claim it will serve urban markets any time soon... That's rediculous - as rediculous as saying Tesla will achieve level 5 autonomous driving this decade
noahtallen 6y ago
IMO, the market is rural folks who don’t have fast internet or any other options. Less than 6mbps down is normal in rural Ohio, for example. And there are no other wired options, and existing satellite options aren’t much better. These people can even be just a mile or less outside of a town with internet options, but none of the telecom companies have upgraded capacity in the countryside for years. Obviously, a modern wired connection would be way better, but ISPs are just not willing to invest in the upgraded infrastructure where nearly no one lives. You could lay a mile of fiber to reach only one or two households. (Or less.) Satellite makes a lot of sense here because it can cover the whole area without having to build a ton of infrastructure relative to the population.

jeffnappi 6y ago

Can you point me to your source on the planned constellation total switching bandwidth? I haven't read about that anywhere. I have read that they don't expect to be able to provide high levels of service in dense urban areas. Which does make me question the business model... I think you nailed it that the military and government will be a big customer. Iridium has ~1300ms+ latency and very low data rates, very high costs and the government is their largest customer from what I understand. Airlines are another likely market that would be a significant amount of business (post-covid19). If Starlink can compete against Iridium and the other existing satellite providers on price, they could at a minimum take most of that business.

foobiekr 6y ago
Per satellite max bandwitdh, per SpaceX, at commercialization will be a max of 17gbps [1, sorry for the lame source but it's what I could find at the moment]. Depending on how they're measuring that, the actual throughout could be substantially less than that. Given that their FCC filing was 20gbps per satellite, I think that's the raw, uncorrected bit rate, not the corrected bitrate which will be lower. A coworker and I did a back-of-envelope guess based on their power budget (switching capacity in bandwidth and pps and power are basically directly tied, ditto capacity and spectrum) and that's about what we arrived at as their upper bound (imho, not even half of that will be achievable). Note that this is a double-counted metric (bps total is bps_in+bps_out) for actual traffic volume. So let's go with that. The current SpaceX fantasy business model is to build out a constellation of 30,000 satellites, but near term not even close to that. Being very generous, let's credit them with an incredible 10k satellites, pretend they are not power constrained and use the full 20gbps allotted to them, and you get a total of 200k gbps total capacity for the constellation, uncorrected, assuming ample power, or on a good day 125Tbps for real data. Using a somewhat older core switch, the Juniper QFX10016, can do line rate switching for 480 100GE ports; 48Tbps in a single 21RU chassis from 2017. [2] These are commonly deployed in pops. The next generation will be doing that for 400GE ports in the same form factor, 192Tbps. A Cisco 8818, which is shipping right now, does 260Tbps per chassis in a 33RU form factor. The point is, anyone claiming we could serve even a large metro with a single one of these switches, let alone planet earth, would be laughed at, yet here we are. A given nation-scale network will have hundreds of routers with these kinds of capacities. I also don't think Starlink will get anywhere near the estimates I'm using here. [1] https://www.teslarati.com/spacex-starlink-satellite-upgrade-more-bandwidth-more-beams/ https://www.teslarati.com/spacex-starlink-satellite-upgrade-... [2] https://www.juniper.net/documentation/en_US/release-independent/junos/topics/topic-map/qfx10016-system-overview.html#power-supplies https://www.juniper.net/documentation/en_US/release-independ... [3] https://www.cisco.com/c/en/us/products/collateral/routers/8000-series-routers/datasheet-c78-742571.html https://www.cisco.com/c/en/us/products/collateral/routers/80... edit: moved the [2] to the right place. edit damage.
- jeffnappi 6y ago
  Thanks for the data points :) I think something to consider here is that there is an iterative development process taking place with the satellites. Given the stated lifespan of the satellites is 3-4 years, they likely expect to see additional innovation over time and are gambling that the network will be improved over time, and may one day actually be capable of serving urban areas. This seems like a wise approach and appears to work well for both SpaceX and Tesla. The most viable use case in the near term is clearly to serve remote and rural areas which is admittedly a small market. There is a chance you might see governments subsidize service to rural areas that are underserved. I also imagine there will be a number of marine and aviation customers - though that's likely even smaller than the rural market. But yeah, I'm not one of the people who will claim it will serve urban markets any time soon... That's rediculous - as rediculous as saying Tesla will achieve level 5 autonomous driving this decade

noahtallen 6y ago

IMO, the market is rural folks who don’t have fast internet or any other options. Less than 6mbps down is normal in rural Ohio, for example. And there are no other wired options, and existing satellite options aren’t much better. These people can even be just a mile or less outside of a town with internet options, but none of the telecom companies have upgraded capacity in the countryside for years. Obviously, a modern wired connection would be way better, but ISPs are just not willing to invest in the upgraded infrastructure where nearly no one lives. You could lay a mile of fiber to reach only one or two households. (Or less.) Satellite makes a lot of sense here because it can cover the whole area without having to build a ton of infrastructure relative to the population.