SpaceX Internet Satellite Network: Starlink

[Cosmacelf]

Ok thanks. So roughly average 2 Mbps would be 250-500Mb/s for the end user?

The end user could get 1000 Mbps. There isn't a direct linear correlation between average access network uplink bandwidth and individual access network bandwidth. Regardless, seems that Starlink, at least initially, is targeting access bandwidth around 150 Mbps per user.

To give you a simple example. If I had 700 customers, and they each decided to download a 1 gigabyte file from a high bandwidth server (that could deliver 1 gigabit/second), it would take about 10 seconds (1 gigabyte is 8 gigabits and then you have overhead). If the only network traffic was each customer downloading that file, each one in turn doing it 10 seconds apart from each other, the access network's uplink (the ISP) demand would only be 1 Gbps since that's all the download traffic being asked for at any one point in time.

Conversely if all 700 customers decided to download that file at the exact same time, the access network's uplink would be swamped and each user would find their download took longer than 10 seconds.

In the real world, most people are browsing the Internet and that might put a 5 Mbps load for the 1/10 second the page contents download (that would be 500K bits, or about 50K bytes, a not unusual payload for a web page these days, assuming no caching, and of course, after the first page on the site, future pages would be a lot less since the browser has now cached all the javascript libraries, fonts, etc. that the site uses). So on average, the ISP doesn't see that much bandwidth demand even if each user has a 1000 Mbps connection.

 

[Matias]

The end user could get 1000 Mbps. There isn't a direct linear correlation between average access network uplink bandwidth and individual access network bandwidth. Regardless, seems that Starlink, at least initially, is targeting access bandwidth around 150 Mbps per user.

To give you a simple example. If I had 700 customers, and they each decided to download a 1 gigabyte file from a high bandwidth server (that could deliver 1 gigabit/second), it would take about 10 seconds (1 gigabyte is 8 gigabits and then you have overhead). If the only network traffic was each customer downloading that file, each one in turn doing it 10 seconds apart from each other, the access network's uplink (the ISP) demand would only be 1 Gbps since that's all the download traffic being asked for at any one point in time.

Conversely if all 700 customers decided to download that file at the exact same time, the access network's uplink would be swamped and each user would find their download took longer than 10 seconds.

In the real world, most people are browsing the Internet and that might put a 5 Mbps load for the 1/10 second the page contents download (that would be 500K bits, or about 50K bytes, a not unusual payload for a web page these days, assuming no caching, and of course, after the first page on the site, future pages would be a lot less since the browser has now cached all the javascript libraries, fonts, etc. that the site uses). So on average, the ISP doesn't see that much bandwidth demand even if each user has a 1000 Mbps connection.

Ok. But you wrote

“ISP delivering gigabit speeds will find their uplink demand around 4 Mbps to 8 Mbps per customer. I chose a slightly degraded number of 2 Mbps for my calcs since that number is still going to be a whole lot better than geo sat service, so rural customers will still be happy with it.”

and since you used average number 2 Mbs (instead of 4-8), it should be slower than an ISP delivering gigabit speed? I’m just trying to understand

Of course I understand that if there is no capping, with good luck speed could be better. With bad luck it could be worse.

 

[Cosmacelf]

Of course I understand that if there is no capping, with good luck speed could be better. With bad luck it could be worse.

Capping is a different concept. Capping is where an ISP limits the total amount of data downloaded over a month. So you can have a gigabit connection, yet the total amount of data downloaded might be capped at 1 TB or something. It is used to prevent people from using their consumer link to serve other customers (as in reselling the link), or using it for an intensive business purpose.

 

[Matias]

Capping is a different concept. Capping is where an ISP limits the total amount of data downloaded over a month. So you can have a gigabit connection, yet the total amount of data downloaded might be capped at 1 TB or something. It is used to prevent people from using their consumer link to serve other customers (as in reselling the link), or using it for an intensive business purpose.

Ok. I meant limiting the top speed.

Still one question; why would Starlink’s customers’ average bandwidth use be less than wired isp users? If users are similar, their average bandwidth use should be the same?

If wired isp customer downloads a file in 10 seconds and then is 10 seconds idle and Starlink’s customer downloads that same file in 20 seconds, their average bandwidth use is the same during that 20 seconds period.

 

[Cosmacelf]

Ok. I meant limiting the top speed.

Still one question; why would Starlink’s customers’ average bandwidth use be less than wired isp users? If users are similar, their average bandwidth use should be the same?

If wired isp customer downloads a file in 10 seconds and then is 10 seconds idle and Starlink’s customer downloads that same file in 20 seconds, their average bandwidth use is the same during that 20 seconds period.

You are correct. I intentionally used a worse service level for Starlink than a wired ISP since I assumed rural Starlink users would be more tolerant of lower service levels which are still much better than what they have now.

 

[Matias]

You are correct. I intentionally used a worse service level for Starlink than a wired ISP since I assumed rural Starlink users would be more tolerant of lower service levels which are still much better than what they have now.

But what would be the mechanism that reduces Starlink users average bandwidth use compared to wired ISP users?

They still could watch Netflix in 4K...

 

[hmcgregoraz]

Hi,

If you really want to do a deep dive on eyeball networks, check out the north american network operators group (nanog, NANOG Homepage)

Over subscription of either bandwidth or connection ports have been standard on consumer internet networks for a very long time. Example, a local dialup ISP back in the 1990s, when they only had ~4000 customers needed a ratio of about 1 modem per 8 customers, or else they would run into periods of time of busy signals. By the time they reached ~20,000 customers, they were able to scale back the ratio to about 1 modem per 13 or 14 customer.

This thread is from 2011, so scale the bandwidth up, but the numbers being talked about were 1,000 100Mbit customers on a 1Gbit backhaul. Contention/Oversubscription maths Even if you scale that up by 10x, your looking at 2,000 100Mbit class customers on a 20Gbit backhaul, which I think is very reasonable.

The larger the amount of consumer bandwidth, both on the customer side, and on the backhaul, the "better" it shares.

Ie if you have 100 100Mbit users on 1Gbit of backhaul, it does not share anywhere near as well as 100 1Gbit users on 10Gbit of backhaul, or even 1000 100Mbit users on 10Gbit of backhaul. I would even state that your network will do better with 1000 1Gbit users on 10Gbit of backhaul vs 100 100Mbit users on 1Gbit of backhaul.

Most user cases (Except streaming and video conf) are bursts, and streaming UHD is about 15-18Mbit/sec when you are doing it, and Video Conf tends to be under 8Mbit/sec even with multiple participants. What you are relying upon, is that while you may have 10,000 users on a backhaul, not all of them will be doing streaming content at the same time, even during "working" hours. Ie most people do not have back to back to back meetings, and remote school does tend to have a schedule of times people are on, and times they are not.

You also have time zones that help, though not many locations will bridge across a time zone with satellite coverage.

Neflix recommends as 25Mbit internet service for UHD, due to sharing and not having the bandwidth available all the time. Internet Connection Speed Recommendations

From a network engineering prospective, I see no issues with having 10,000 users on a satellite.

This connection is a "worst case", it has 100/100 small business fiber, 4 tenants, some web hosting (including nextcloud instances) and mail servers, and a remote backup that replicates data out 3x per day per server, and bandwidth usage average for the last week (with catching up on a backup) is under 6Mbit/in and 6Mbit/out

Looking at some of my tenants:

-Harry

 

[Cosmacelf]

But what would be the mechanism that reduces Starlink users average bandwidth use compared to wired ISP users?

They still could watch Netlix in 4K...

Access network congestion. If the demand is higher than what the access network can provide then each user will get a smaller slice than they wanted. So at peak times, they might not be able to watch 4k. But HD might work and that might just fine for rural users without wired options.

 

[Matias]

Why is the antenna steerable? It is phased array, so it steers beam electronically. I understand that it needs to be horizontal, but it could be adjusted manually.
Are there so large distances between satellite trajectories, that it needs to adjust position to have satellites boresights?

Does the antenna move during operation or only during set up?

 

[bxr140]

It’s not steerable in so far as tracking satellites—near as we (or at least I) know it’s only for optimizing pointing during setup. Certainly it will optimize by latitude/longitude (the latter for covisibility to ground stations) including reoptimization as the constellation grows, but probably also by some recurring diagnostic that identifies obstructions like buildings and foliage, potentially seasonally.

 

[mongo]

Why is the antenna steerable? It is phased array, so it steers beam electronically. I understand that it needs to be horizontal, but it could be adjusted manually.
Are there so large distances between satellite trajectories, that it needs to adjust position to have satellites boresights?

Does the antenna move during operation or only during set up?

In addition to @bxr140 's post, the beam is most focused when centered, so tilting the dish toward the area with best coverage improves performance. Pages 11 and 12 of the technical pdf have more data about this.

 

[Cosmacelf]

In addition to @bxr140 's post, the beam is most focused when centered, so tilting the dish toward the area with best coverage improves performance. Pages 11 and 12 of the technical pdf have more data about this.

oooh. I missed that. Do you have a link to the technical pdf?

 

[mongo]

oooh. I missed that. Do you have a link to the technical pdf?

The link I found doesn't put the .pdf extension on it. I attached the file to a previous post: SpaceX Internet Satellite Network: Starlink

 

[BioSehnsucht]

Very in depth teardown and discussion of how the terminal (probably - he admits he's not an RF expert) works.

 

[Cosmacelf]

Um, how are people doing tear downs? Surely you don’t buy the dish when you do a beta? Presumably, it is still owned by SpaceX?

 

[bkp_duke]

Um, how are people doing tear downs? Surely you don’t buy the dish when you do a beta? Presumably, it is still owned by SpaceX?

$499, pretty sure you have to "buy" the dish for the public beta.

 

[Cosmacelf]

$499, pretty sure you have to "buy" the dish for the public beta.

But the dishes are obviously subsidized. You'd have to pay for many months if not years of service for SpaceX to break even. I guess you'd have to read the terms of service (which I don't have) too see what restrictions, penalties, contract length, etc. SpaceX is imposing.

 

[bkp_duke]

But the dishes are obviously subsidized. You'd have to pay for many months if not years of service for SpaceX to break even. I guess you'd have to read the terms of service (which I don't have) too see what restrictions, penalties, contract length, etc. SpaceX is imposing.

Subsidized or not, I'm 99.9% certain if you cancel the service, you don't have to return the dish. So if you don't have to return it, you effectively own it and can tear it down if you want.

Would love for an actual beta tester to look through the T&C and verify this.

 

[Zaxxon]

Am I the only one considering the somewhat hair-brained idea of opting for Starlink beta despite living in the burbs and having solid broadband access?

I understand it's not likely to top Comcast, but would much rather my $$ go to SpaceX rather than Comcast (who just expanded data caps and announced another price increase).

Currently I pay $90/mo for 1 Gbps/40Mbps, which in reality is around 650 Mbps/35Mbps. I'm right on the edge of Starlink's beta area and expect I'll be in range come the January/February beta expansion. I can stomach paying the same for lower speeds in return for the cool tech factor. Could even keep a basic tier of Comcast service as a backup.

Am I alone?

 

[bkp_duke]

Am I the only one considering the somewhat hair-brained idea of opting for Starlink beta despite living in the burbs and having solid broadband access?

I understand it's not likely to top Comcast, but would much rather my $$ go to SpaceX rather than Comcast (who just expanded data caps and announced another price increase).

Currently I pay $90/mo for 1 Gbps/40Mbps, which in reality is around 650 Mbps/35Mbps. I'm right on the edge of Starlink's beta area and expect I'll be in range come the January/February beta expansion. I can stomach paying the same for lower speeds in return for the cool tech factor. Could even keep a basic tier of Comcast service as a backup.

Am I alone?

I would consider that in your shoes as well.

Sadly, I'm out of Starlink service area. But I do have a sweet deal with AT&T Fiber for 1/1Gbs, unlimited, lifetime contract for $70/mo.