SpaceX Internet Satellite Network: Starlink

[scaesare]

If it goes up dramatically, wouldn't that indicate the link limit is 17Mbps? Otherwise, they would need to be hiding retries or adjusting the link parameters based on traffic as packets 'should' be independent.

I'd tend to think so....

Unless you are talking about boosting both the link rate and the iperf3 rate?

No, I was just wandering what we'd see iperf3 was run at a higher rate. I'm not used to seeing a 15% packet loss as being terribly acceptable, especially for media where it's audible/visible with UDP. That having been said, I've never really measured over a cell connection (more a LAN dude... although I see there's an iperf port for iOS, hmmm...), so maybe that's acceptable, and the FEC and codecs are such that it's not as objectionable as I'd expect...

So... the test would be interesting to see.

 

[mongo]

No, I was just wandering what we'd see iperf3 was run at a higher rate. I'm not used to seeing a 15% packet loss as being terribly acceptable, especially for media where it's audible/visible with UDP. That having been said, I've never really measured over a cell connection (more a LAN dude... although I see there's an iperf port for iOS, hmmm...), so maybe that's acceptable, and the FEC and codecs are such that it's not as objectionable as I'd expect...

So... the test would be interesting to see.

Yeah, that is a super high rate for normal situations. There is a whole science for bit error rates based on signal to noise, bandwidth, and all that stuff. Since Starlink is using standard cellular packets, they can't compensate with extra error correction.

Were it not for that first second of data, I'd more easily chalk it up to link margin. If they run the test at <15Mbps, that could also show where the issue is.

The test was run in reverse mode, so an iOS version would theoretically give the same results since it's the other end doing the transmitting (based on my zero knowledge of the internals).

 

[scaesare]

Yeah, that is a super high rate for normal situations. There is a whole science for bit error rates based on signal to noise, bandwidth, and all that stuff. Since Starlink is using standard cellular packets, they can't compensate with extra error correction.

Were it not for that first second of data, I'd more easily chalk it up to link margin. If they run the test at <15Mbps, that could also show where the issue is.

The test was run in reverse mode, so an iOS version would theoretically give the same results since it's the other end doing the transmitting (based on my zero knowledge of the internals).

I had mentioned the iOS port in that I could grab it and run some tests over the terrestrial cell network to see what the packet loss there looks like for comparison and to see if my concern over that loss rate was justified.

 

[JB47394]

New Jersey man charged with trafficking 675 SpaceX Starlink terminals

Elon Musk's company SpaceX makes the Starlink terminals that were seized during a police stop. More than $400,000 of those terminals were fraudulently bought.

www.cnbc.com

The total value of the 675 fraudulently purchased Starlink terminals that police subsequently learned had been shipped to the Lawrence Township address is about $400,000, police said.

 

[scaesare]

New Jersey man charged with trafficking 675 SpaceX Starlink terminals

Elon Musk's company SpaceX makes the Starlink terminals that were seized during a police stop. More than $400,000 of those terminals were fraudulently bought.

www.cnbc.com

Given these are useless without service, and SpaceX would be able to determine serial # and location once activated, trafficking/using stolen terminals seems dumb (hard to believe there are dumb criminals, I know...)

 

[mongo]

Direct to mobile is working indoors... (at least sometimes)
h/t Sawyer Merritt
SpaceX: Cellular Starlink System Works on iPhone, Pixel, Galaxy Devices

 

[scaesare]

I continue to be amazed that works...

 

[mongo]

Latency reduction update

https://api.starlink.com/public-files/StarlinkLatency.pdf

 

[JB47394]

Latency reduction update

https://api.starlink.com/public-files/StarlinkLatency.pdf

Apparently, p50 and p99 are percentiles, but I'm not sure which percentiles. Is p50 the latency that the best performing half of customers saw?

 

[mongo]

Apparently, p50 and p99 are percentiles, but I'm not sure which percentiles. Is p50 the latency that the best performing half of customers saw?

"To measure Starlink’s latency, we collect anonymized measurements from millions of Starlink routers every 15 seconds. These 15 second average latencies are then used to calculate the median and worst-case latency. The median (50th percentile or p50) refers to the point where half of the latency measurements are below that number and the other half are above. The worst-case latency, or 99th percentile, is defined as the place where 99% of measurements are better than the point."

 

[bmah]

I'd tend to think so....




No, I was just wandering what we'd see iperf3 was run at a higher rate. I'm not used to seeing a 15% packet loss as being terribly acceptable, especially for media where it's audible/visible with UDP. That having been said, I've never really measured over a cell connection (more a LAN dude... although I see there's an iperf port for iOS, hmmm...), so maybe that's acceptable, and the FEC and codecs are such that it's not as objectionable as I'd expect...

So... the test would be interesting to see.

Picking a random reply...sorry for the delay, it's been a busy week and I'm only now catching up on TMC stuff.

We've been batting this around at work a little bit because making well-performing end-to-end network paths is a big part of what we do. We thought it'd be interesting to see UDP tests at a lowered rate (< 20Mbps) to see what that does to the loss and "goodput". Like one of my coworkers said it'd be much more useful to get (for example) 10Mbps with minimal loss than 20Mbps with 15% packet loss. (For scientific data transfers, which are unlikely to be to or from mobile phones, 15% packet loss basically makes a path unusable. Fortunately our Starlink service uses normal dishes.)

Other thing...many applications on the Internet run over TCP (including HTTP and things like it), which as many of you know, needs two-way communication. So...what's the backchannel like (phone to satellite)?

Bruce.

PS. It's awesome when I can legitimately include reading TMC as a part of work.

 

[scaesare]

Latency reduction update
View attachment 1025858
https://api.starlink.com/public-files/StarlinkLatency.pdf

Lot of improvement in just the last 2 months..

 

[scaesare]

Picking a random reply...sorry for the delay, it's been a busy week and I'm only now catching up on TMC stuff.

We've been batting this around at work a little bit because making well-performing end-to-end network paths is a big part of what we do. We thought it'd be interesting to see UDP tests at a lowered rate (< 20Mbps) to see what that does to the loss and "goodput". Like one of my coworkers said it'd be much more useful to get (for example) 10Mbps with minimal loss than 20Mbps with 15% packet loss. (For scientific data transfers, which are unlikely to be to or from mobile phones, 15% packet loss basically makes a path unusable. Fortunately our Starlink service uses normal dishes.)

Other thing...many applications on the Internet run over TCP (including HTTP and things like it), which as many of you know, needs two-way communication. So...what's the backchannel like (phone to satellite)?

Bruce.

PS. It's awesome when I can legitimately include reading TMC as a part of work.

Good question. Back in the dark ages of the late mid-late 90's we were doing geosync sat IP (with atrocious latency), and were one of the first to use the birds for backchannel (rather than dial-up)... and it was in the 10's of Kbps... and surprisingly that really was not an issue, due to the async nature of stuff like HTTP requests being ~10-100X smaller than the replies... but then again home video conferencing wasn't a thing then either... would like to see what it looks like with Starlink.

 

[JB47394]

SpaceX gets E-band radio waves to boost Starlink broadband

SpaceX has secured conditional approval to use extremely high-frequency E-band radio waves to improve the capacity of its low Earth orbit Starlink broadband constellation.

spacenews.com

 

[Cosmacelf]

SpaceX gets E-band radio waves to boost Starlink broadband

SpaceX has secured conditional approval to use extremely high-frequency E-band radio waves to improve the capacity of its low Earth orbit Starlink broadband constellation.

spacenews.com

"The FCC continues to defer action over whether to allow SpaceX to deploy the other 22,500 satellites in its proposed Gen2 constellation."

I wonder if SpaceX is pushing hard on this or whether their current # of satellites is good enough for now. Seems that they have enough bandwidth in the sky at the moment?

 

[mongo]

"The FCC continues to defer action over whether to allow SpaceX to deploy the other 22,500 satellites in its proposed Gen2 constellation."

I wonder if SpaceX is pushing hard on this or whether their current # of satellites is good enough for now. Seems that they have enough bandwidth in the sky at the moment?

In a rational world, the FCC delaying approval means the deadline for launching the satellites also shifts which provides more opportunity for using Starship and/or full sized V2s versus minis.

 

[jerry33]

"The FCC continues to defer action over whether to allow SpaceX to deploy the other 22,500 satellites in its proposed Gen2 constellation."

I wonder if SpaceX is pushing hard on this or whether their current # of satellites is good enough for now. Seems that they have enough bandwidth in the sky at the moment?

Just part of the Administration push against Elon's companies. I really dislike his current political stand, but I get why he's ticked-off at the current Administration.

 

[JB47394]

Just part of the Administration push against Elon's companies.

Or the FAA doing their job. Placing another twenty two thousand satellites into low Earth orbit is nothing to rubber stamp. And that's on top of all the other constellations that companies want to put up.

 

[Cosmacelf]

Musk has massive conflicts of interest between SpaceX and Tesla. With Tesla he wants to sell cars globally, and China is a critical market. How is China going to react to another Musk company building a vast network of spy satellites that will of course be spying on China, and basically every other country in the world?

So, have you seen any evidence of Elon not being able to handle these supposed conflicts of interest? I have seen no impact to Tesla from Elon running SpaceX.

 

[ecarfan]

So, have you seen any evidence of Elon not being able to handle these supposed conflicts of interest? I have seen no impact to Tesla from Elon running SpaceX.

I have, but that is off topic for this thread, thanks. Apologies to the moderators for my post potentially taking this thread off topic.