Starlink : Speed Discussion

[RedLineTrain]

It appears that the speed spigot has been turned on a bit more.  Today, so far we are seeing as high as 91 Mbps down and 27 Mbps up.

https://testmy.net/host-history/spacex_starlink

[RedLineTrain]

And now it looks like the speeds have taken another step upward.  As high as 105 Mbps and often 100+ Mbps.

https://testmy.net/host-history/spacex_starlink

[RedLineTrain]

And now 122 Mbps...

(Sorry for the rapid-fire posts, but SpaceX must have really made a breakthrough on speeds.)

https://testmy.net/host-history/spacex_starlink

[king1999]

And now 122 Mbps...

(Sorry for the rapid-fire posts, but SpaceX must have really made a breakthrough on speeds.)

https://testmy.net/host-history/spacex_starlink

You may need to post screenshots next time. They all went down now.

[Nomadd]

 I'm not exactly sure what this thread is good for. Anybody who didn't sleep through 3rd grade math for two years, and has any idea how ISPs work knows that these early tests have pretty much nothing to do with a mature, highly subscribed system. And I've been seeing some pretty nonsensical proclamations by pretend experts.
 
 I use to test systems by downloading files from some oddball site called nasaspaceflight.com. It had the most reliably consistent service I knew. Then it got big and I found out what Chris did to people who wasted his bandwidth, so I abused some other site instead.

[RedLineTrain]

I'm not exactly sure what this thread is good for.

To track Starlink's speed progress and to discuss it.  Who knows, we may see some tests that diverge from the expectations of the pretend experts.  Or we may have page after page of fruitless arguments over what is written in the tea leaves.

Seems in keeping with the rest of the site.

[RedLineTrain]

And now 122 Mbps...

(Sorry for the rapid-fire posts, but SpaceX must have really made a breakthrough on speeds.)

https://testmy.net/host-history/spacex_starlink

You may need to post screenshots next time. They all went down now.

The 122 Mbps test is still in the log, so no need for a screenshot.  But if you want to learn more about that particular test, below is a link to the test.

https://testmy.net/db/Hwm-K0S7E

[abaddon]

I'm not exactly sure what this thread is good for. Anybody who didn't sleep through 3rd grade math for two years, and has any idea how ISPs work knows that these early tests have pretty much nothing to do with a mature, highly subscribed system. And I've been seeing some pretty nonsensical proclamations by pretend experts.
 
 I use to test systems by downloading files from some oddball site called nasaspaceflight.com. It had the most reliably consistent service I knew. Then it got big and I found out what Chris did to people who wasted his bandwidth, so I abused some other site instead.

Sure, and Starship prototype testing has nothing to do with a mature, highly refined system.  Should we not care how early Sharship testing goes?

I love your posts, but this one is disappointing and does not contribute value to this thread.  And I'm a professional in high-speed low-latency networked applications, so I have at least some inkling of what I'm talking about.

[launchwatcher]

From the starlink FCC filing thread, relevant to observed performance during the beta test:

SpaceX's beta testing uses conditions
designed to support on-going optimization and testing of the network that make network
performance measurements worse, not better. For example, all the user terminals were
configured to transmit debug data continuously, even if the beta customer didn't have any
regular internet traffic, forcing every terminal to continuously utilize the beam.

Moreover, these results are based on beta-test software frame grouping settings that
do not yet reflect performance using the software designed to optimize performance for
commercial use. Until recently, the network had been grouping user terminals in groups
of 8 per radio-frame, instead of the 20 terminals per radio-frame the system supports. This
operating choice is to support on-going optimization and testing of the network but has the
consequence of introducing 2.5 times longer delay between radio-frames for a given user
in a fully loaded cell, corresponding to the smaller group sizes. Importantly, this software
feature has just been enabled and is specifically designed to optimize speeds in highly
populated cells, increasing throughput by approximately 2.5 times.

In addition to the datapoints representing SpaceX's aggregate performance, SpaceX
analyzed the last week of measurements for a community of 30 high-usage customers. As
shown in Figure 1, these measurements, totaling 1,048,576 datapoints, indicated a
95th percentile latency of 42 ms and percentile latency of 30 ms between end users and
the point of presence connecting to the Internet. These measurements confirm the SpaceX
network is capable of allocating resources efficiently such that latency remains consistent
whether the measurement point is the overall network or individual groups of customers.

[Mandella]

From the starlink FCC filing thread, relevant to observed performance during the beta test:

SpaceX's beta testing uses conditions
designed to support on-going optimization and testing of the network that make network
performance measurements worse, not better. For example, all the user terminals were
configured to transmit debug data continuously, even if the beta customer didn't have any
regular internet traffic, forcing every terminal to continuously utilize the beam.

Moreover, these results are based on beta-test software frame grouping settings that
do not yet reflect performance using the software designed to optimize performance for
commercial use. Until recently, the network had been grouping user terminals in groups
of 8 per radio-frame, instead of the 20 terminals per radio-frame the system supports. This
operating choice is to support on-going optimization and testing of the network but has the
consequence of introducing 2.5 times longer delay between radio-frames for a given user
in a fully loaded cell, corresponding to the smaller group sizes. Importantly, this software
feature has just been enabled and is specifically designed to optimize speeds in highly
populated cells, increasing throughput by approximately 2.5 times.

In addition to the datapoints representing SpaceX's aggregate performance, SpaceX
analyzed the last week of measurements for a community of 30 high-usage customers. As
shown in Figure 1, these measurements, totaling 1,048,576 datapoints, indicated a
95th percentile latency of 42 ms and percentile latency of 30 ms between end users and
the point of presence connecting to the Internet. These measurements confirm the SpaceX
network is capable of allocating resources efficiently such that latency remains consistent
whether the measurement point is the overall network or individual groups of customers.

Thanks for crossposting that. I haven't been following the filing thread.

So, basically Starlink is in debug mode right now, as everyone should have assumed.

I do think that network congestion is unavoidably going to bring performance down dramatically when everyone logs in at once for whatever reason, but Starlink is still going to be head and shoulders above regular satellite, or gods forbid rural cellular...

[su27k]

Speedtest article about Starlink speeds: Starlink: Bridging the Digital Divide or Shooting for the Stars?

In both the U.S. and Canada, Starlink provided competitive or better service at the minimum (25/3/100) tier. In the U.S., 86.7% of Starlink users met this threshold, compared with 83.2% of those on all other fixed broadband providers. Although the FCC’s criteria don’t apply north of the border, 85.6% of Canadian Starlink users met the Minimum threshold, compared with 77.8% for all other providers. Starlink showed a smaller proportion of users meeting the baseline and above baseline tiers than all other providers combined.

Given this data, it's safe to say Starlink could be a cost-effective solution that dramatically improves rural broadband access without having to lay thousands of miles of fiber.

[Hog]

I'm not exactly sure what this thread is good for. Anybody who didn't sleep through 3rd grade math for two years, and has any idea how ISPs work knows that these early tests have pretty much nothing to do with a mature, highly subscribed system. And I've been seeing some pretty nonsensical proclamations by pretend experts.
 
 I use to test systems by downloading files from some oddball site called nasaspaceflight.com. It had the most reliably consistent service I knew. Then it got big and I found out what Chris did to people who wasted his bandwidth, so I abused some other site instead.

Sure, and Starship prototype testing has nothing to do with a mature, highly refined system.  Should we not care how early Sharship testing goes?

I love your posts, but this one is disappointing and does not contribute value to this thread.  And I'm a professional in high-speed low-latency networked applications, so I have at least some inkling of what I'm talking about.

Emphasis mine.
Nomad's injection of humour gave this thread increased value.

[su27k]

Some great speed statistics for Starlink: How Starlink’s Satellite Internet Stacks Up Against HughesNet and Viasat around the Globe

The satellite internet race is heating up, with more competitors serving more areas than even a quarter ago. We’re back with a broader look at internet network performance for satellite providers across the globe based on Q2 2021 data from Speedtest Intelligence™.

Quick summary from subtitles of the article:

Starlink speeds beat competitors in the U.S., can’t top fixed broadband

Canada: Starlink beats fixed broadband download speeds

France: Where the fastest Starlink download speeds are

Germany: Starlink far surpasses local fixed broadband

New Zealand: Starlink is faster than fixed broadband

United Kingdom: Starlink beats fixed broadband providers

It's also interesting to see that US has the best fixed broadband speed among western countries, quite contrary to the picture media is painting about how bad US broadband infrastructure is.

[RonM]

It's also interesting to see that US has the best fixed broadband speed among western countries, quite contrary to the picture media is painting about how bad US broadband infrastructure is.

US broadband is only available in urban areas and to people who can afford it. The US is a vast country with rural communities unable to get broadband because it would be too expensive to setup via traditional cable or fiberoptic cable. Tens of millions of Americans cannot access broadband. Starlink is designed to tap into this large market.

[su27k]

Starlink - The Billionaire Elon Musks’ Dream for Broadband from Space is Still in Infancy Yet Gaining Popularity Among Consumers

Speedcheck carried out a series of speed tests over the last couple of months to measure the performance of Starlink, which is now in the beta testing phase. This article looks into what Starlink is doing and how it differentiates itself from other satellite, fixed and wireless broadband service providers.

The results are convincing, and we have seen modest figures in the range of 50.5 Mbps download speed, 14 Mbps upload speed, and 52.5 ms latency. We have also carried out tests in the US and Canada and found identical results. The US's median upload and download speeds were 13 Mbps and 50 Mbps, respectively, with 57ms latency. And the median download and upload speeds in Canada were 49 Mbps and 14 Mbps, respectively, with 52 ms latency.

[Bob Niland]

RonM: … The US is a vast country with rural communities unable to get broadband because it would be too expensive to setup via traditional cable or fiberoptic cable. Tens of millions of Americans cannot access broadband. Starlink is designed to tap into this large market. …

I'm one of those customers. Here's the back-story, and some crystal ball gazing.

Dial-up was never faster than 28K here. ISDN, DSL, cable & fiber never made it here. The GEO services are still the sad joke that they always were.

What we do have is two providers on the old Motorola Canopy terrestrial RF tech, and/or 4G cellular IP (5G may be too weak here). The RF tech hasn't seen any real upgrades in a decade. It's a footnote business for the regional ISPs, who offer mainly fiber & cable in the larger communities.

For anyone with line-of-site to the terrestrial RF towers, they offer 2Mbps & 5Mbps at $55 or $70/mo. It's reasonably reliable, with a couple of outages a year (during which we fallback to cellular IP). For $25 more, Starlink represents an order of magnitude jump in performance, and I expect many to jump on it when Sx opens it to general availability here.

Because Starlink will be so appealing to so many customers on the existing RF, I'm frankly expecting RF subscriptions to sink, and the regional ISPs to deprecate and eventually abandon the RF tech. Thus, although I don't really need the extra bandwidth, I'm planning to switch to Starlink on my schedule, and not wait until the RF goes the way of landline.

[Redclaws]

It's also interesting to see that US has the best fixed broadband speed among western countries, quite contrary to the picture media is painting about how bad US broadband infrastructure is.

US broadband is only available in urban areas and to people who can afford it. The US is a vast country with rural communities unable to get broadband because it would be too expensive to setup via traditional cable or fiberoptic cable. Tens of millions of Americans cannot access broadband. Starlink is designed to tap into this large market.

The US also has pretty high broadband costs.  Our broadband infrastructure certainly isn’t terrible but it’s got gaps and is deeply non-competitive in most regions.

[envy887]

Starlink - The Billionaire Elon Musks’ Dream for Broadband from Space is Still in Infancy Yet Gaining Popularity Among Consumers

Speedcheck carried out a series of speed tests over the last couple of months to measure the performance of Starlink, which is now in the beta testing phase. This article looks into what Starlink is doing and how it differentiates itself from other satellite, fixed and wireless broadband service providers.

The results are convincing, and we have seen modest figures in the range of 50.5 Mbps download speed, 14 Mbps upload speed, and 52.5 ms latency. We have also carried out tests in the US and Canada and found identical results. The US's median upload and download speeds were 13 Mbps and 50 Mbps, respectively, with 57ms latency. And the median download and upload speeds in Canada were 49 Mbps and 14 Mbps, respectively, with 52 ms latency.

Speedcheck runs on the user device, which is really not a great way to evaluate the Starlink network speed. WiFi obstructions or congestion on the local network can easily drop speeds to the user device by a huge amount, so that it doesn't matter whether the ISP is providing 1000 Mbps or 30 Mbps to the modem, because only 30 Mbps is getting through the WiFi. Network evaluations should be done only on systems where the local network is known to not be the slowest link.

The Starlink app run both a speed check to the device, and a speed check to the router. These are the same if I'm standing right by the router (usually 80-130 Mbps), but the to-device speed varies considerably as I move around the house, often below 30 Mbps.

The average over the last 2 months or so at https://starlinkstatus.space/ is about 150 Mbps, aggregated from about 100 users. I think those users, having gone through the effort to install a program to do automated speed checks, have also mostly made sure that the device doing the speed checks is not on a slower local network.

[macpacheco]

However, as can be seen from the snip, speedtest thinks the user-server are < 50 miles apart.

My emphasis.  The ways that speedtest.net uses to determine the user's geographical location are however not very reliable for a brand new ISP which is only in beta-test with a small number of users spread all over the US.  The beta-testers may all be within the same small range of IP addresses, even though they are spread from Los Angeles to Brownsville to Boston to Seattle.  Speedtest, and the geolocation services like MaxMind, are very likely to clump together all current Starlink users to a single location today.

This is especially likely if Starlink have not yet built out their infrastructure to put people in different IP ranges depending on their location (since there are so few users at the moment).  Or have not yet established peering with enough number of networks in enough number of places around the US.  Or simply have not kicked MaxMind and their ilks hard enough to get correct information into the geolocation databases...

(If you read the NANOG, North American Network Operators' Group, email list, a recurring problem people have is that Netflix/Disney+/Hulu/et.c suddenly thinks their network is in outer Mongolia instead of in the US, and refuses to serve content to their customers.)

It's way too early to draw any conclusions about how well the Starlink network will work when they open up to real customers.

This assumes the ISP can and will sort out the IPs in neat ranges and register them to different locations.
I work for a regional ISP, all ips are registered to the same location, but they are deployed over a 300 mile radius.
I assume medium nationwide ISPs simply can't organize their ips by state, that would waste way too many.
I assume Starlink simply won't be able to do anything more than clump ips by continent, if at all.

[su27k]

Starlink Expands but Q3 2021 Performance Flattens in Some Areas

Satellite internet is making headlines across the globe as Starlink continues to launch service in new countries and Viasat plans to acquire Inmarsat. We’re here to check in on our ongoing series on satellite internet performance around the globe with fresh data from Q3 2021 to see if Starlink’s performance is holding up and how satellite internet compares to fixed broadband in 12 countries.

Quick summary from subtitles of the article:

United States: Starlink fastest, speeds decreasing

Australia: Starlink outperformed Viasat and fixed broadband average

Belgium: Starlink dramatically faster than fixed broadband

Canada: Starlink performance relatively flat

France: Starlink speeds decreasing, remains faster than fixed broadband average

Germany: Starlink faster than country average for fixed broadband

New Zealand: Starlink slowed slightly, remained faster than fixed broadband average

United Kingdom: Starlink nearly twice as fast as fixed broadband average