Hacker News Reader

Comment by geerlingguy | Hacker News Reader

Honestly when I first got the dish, and had it in an open field, that was the case—but now the momentary dropouts between satellites are less than 1-2 seconds.

They had 1000 or so satellites when I first started testing, and there are now something like 1600 or so. Most of the time, I don't even notice when it switches satellites.

If you were doing some more real-time work or extremely latency-sensitive operations, then yes, you need to stick to a different type of connection. But it's really seamless now, compared to even a few months ago.

Most of the software I used either showed no sign of the dropout, or at worst would freeze a frame or show a loading indicator for a brief moment before getting back to normal.

Online multiplayer gaming and/or streaming are the main areas where I'd have to not recommend Starlink for now.

View on HN

I'm thinking of using this for a backup connection if FIOS fails... we'd switch our office over to starlink... currently we switch to comcast and it's basicaly unsable... so wonder if starlink would be good for ssh connections etc...

I work from home on Starlink. My SSH sessions work just fine. When I first got the service, it was dropping quite a lot, but even then SSH worked pretty well. But these days, if there are any drops, I can't tell. The improvements in the last few months have been massive. I am even about to drop my land line, because WiFi calling on my cell phone has been nearly perfect. It works better than my cell phone does when I am in a good service area.

mosh instead of ssh helps for unreliable connections

When I connect to remote servers over SSH, I usually start a screen (another alternative is tmux) session and then perform my administrative and diagnostics actions in that session. This way, even if I get disconnected, I can reconnect and join that running session. There are other benefits to a screen multiplexer - such as multiple remote users being able to attach to the same session, a joy when a distributed team needs to diagnose an issue or watch some actions together.

You may also want to look into mosh [1] as it will handle dropouts and continue where ssh would just fail on its own. Also has local echo for points of high latency.

Was a real life saver on some of the terrible sat connections of yesteryear.

1. https://mosh.org/

Mosh is great, but if you don't need the predictive terminal emulation, then you can try Eternal Terminal, which is easier to get running due to not needed an UDP port open.

How does it compare to autossh?

autossh just re-reestablishes your ssh connection. It doesn't tolerate packet loss very well or have local echo like mosh.

mosh does require UDP ports, though, which might be problematic with some firewalls or proxies.

Combine with mosh and you can even keep editing while the connection blips!

yeah i use screen for everything too... still there are times when a connection dropping is a bit scary...

LTE makes a lot more sense if you have a decent signal. The cost is lower and speeds are comparable in many areas if you have good hardware.

If you live somewhere with FiOS, Starlink is going to be a non-starter. FiOS requires high density. High density kills Starlink connections.

I suppose Starlink can support a small, negligible, percentage of customers in an urban (or even suburban) area. (Same total number as a rural area.) So, if you think they are going to cutoff registrations as oppose to oversubscribe when they hit that point you may want to sign up right away.

I'm not sure if you are using FiOS as a synonym for FTTH. If so, you might be interested to know I live somewhere with FTTH that is reasonably low-density (~25-30 people per sq km). I think Starlink will be a very serious consideration here.

When I say FiOS, I'm thinking gigabit+ FTTH/FTTD (I'm not 100% sure of the difference). Given that you quoted people per sq km (about 70-84 people per sq. mile), you're probably not in the US. In the US, it's hard to get fiber outside of areas with the high density I'm quoting (at least in my experience).

But yes, that density seems fine for Starlink (although maybe it isn't in the end, who knows yet.)

My understanding is that StarLink is aiming for 100M un-guaranteed in beta. So that's considerably below gigabit.

But I also know that most households with FTTH around here use considerably less than 50M.

So perhaps you meant that gigabit internet kills Starlink. FTTH has some advantages over Starlink. But Starlink also has advantages over FTTH. If you need more than 100M, Starlink is not in the game yet. But if you want some leverage with the local monopoly and have typical usage...

Or, if you snowbird (common case here) and want the same provider on the border of Canada and the border of Mexico...

The total aggregate bandwidth Starlink will have when their entire constellation of satellites is deployed and the number of satellites that will be visible at a given location at a given time place an upper limit on the number of customers they can have even with overselling bandwidth.

In the US their plans are to be able to support something like 5 million customers. That's less than 5% of the number of customers Comcast has.

It really is by design meant for people in areas that do not have terrestrial broadband.

This is exactly right!

I think that 5M is a high number (at least with the current number of satellites and their architecture), specially if they do not apply data caps.

I would say that 1M customers in the US is a more reasonable figure. If they hit these number, I expect the performance to degrade significantly.

I thought something like 20% of America did not have terrestrial broadband. At 5 million this is a giant change, but not as universal as I originally thought.

I wasn't referring to the speed (although obviously fiber has higher speed and reliability, and that will likely never change). I was referring to the fact that it seems like if you have the population density (in the US, at least in my knowledge) to justify a local ISP rolling out fiber to homes, you have a population density that will oversubscribe even the completed Starlink constellation (locally). Starlink just is never going to work in NYC or SF or areas with significantly less density there. Although it may work for a lucky few who manage to enroll before the area is full and locked down.

Some low-density parts of the US have fiber as well thanks to programs like https://www.usda.gov/reconnect

Sure, but it's some for a reason. It's a generality. So is "if you're shopping in Goodwill, you're not driving a Ferrari". Obviously, the recently broke, extremely thrifty, people looking for a collectable or to reclaim something accidentally donated are exceptions.

Maybe that program has covered more than enough people to make it worth talking about, but I didn't see anything like that.

I’d fix the Comcast problem before switching over to something that is pretty much guaranteed to have drop offs at this point in time.

There are also devices that will give you a cellular LTE backup like the U-LTE from UniFi. I’ve heard the plans are a bit pricey.

https://store.ui.com/collections/unifi-accessories/products/...

Yeah we have mifi’s as backup as well but usually only a few people can be on those. But also thinking about dedicated fiber… but honestly none of these seem as cool to me right now as bunch of satellites

For ssh you could use tmux or screen in case you get disconnected

I have five different Wi-Fi APs on my property. The handoff between them results in roughly 40-60ms dropout. That still interrupts audio calls. 1-2 seconds is huge.

Yeah, I am on lots of video and audio calls for work. 1-2 second drops while doing a client presentation are absolutely not ok.

I’m a multiplayer developer and I’m waiting to get a Starlink to test on (I have tested a few times on my moms). I want to makes games tolerate of the latency and instability of Starlink. Mobile networks deal with similar issues so I think it’s possible to make games with Starlink in mind.

I believe these kinds of latency and dropped packet issues be easily simulated on a normal PC without the need for a Starlink connection

Most tools for simulating latency and packet loss locally are a pretty poor model of the kinds of behavior encountered on real networks. In particular, latency due to bufferbloat rather than speed of light delays is extremely important to simulate, but not well-supported by many tools.

I do simulate these conditions, but there’s no replacement for real world condition testing.

exactly. Just use charles proxy (or mitmproxy) and ask it to drop packets/throttle the connection. No need to wait for starlink.

I’m not sure if it’s an endeavour worth taking honestly.. Starlink is still in beta, and once they can use starship to launch the satellites (hopefully somewhere next year) the launch capability will increase to 400 satellites compared to 60 on falcon 9. At that point they will be limited just by the satellites build throughput, so they will reach the 11k coverage very quickly.

I’m already getting a Starlink and I have one available to test on at my parents house so I think it’s worth it. I’m also evaluating it as a way for me to move closer to my family. I’m stuck in Portland because I require really good internet as I’m a 100% remote game dev working on multiplayer games. And the only option in Albany is really bad and expensive Comcast service.

What about an SSH connection? Do you have to reconnect when it switches satellites?

No, the connection seems to be paired up through the ground station, so I wouldn't get disconnected via SSH. Mosh may be a better option if you want to make it feel rock solid though.

> They had 1000 or so satellites when I first started testing, and there are now something like 1600 or so. Most of the time, I don't even notice when it switches satellites.

Enjoy the early adopter moment. Even if they keep increasing the numbers, they will probably move those new satellites in a much wider net to cover more subscribers the second they must show a profit.

They are planning to increase the number of satellites by an order of magnitude. They’re launching extremely fast with Falcon 9, and prepping Starship for launch as well, which has 5-10 times the payload. May see Starlink launches on Starship join Falcon9 within a year or so.

So the opposite is true. They’re likely to massively increase the number of satellites.

If you want to argue the per user bandwidth might be different than for early users, that’s somewhat more plausible. But the number of satellites will increase. They can’t actually significantly change the inclination of the satellites once launched as it takes an insane amount of propellant, and even for solar electric thrusters, so your concern about them moving the satellites to other orbits is very unlikely.

No. They are increasing satellites by an order of magnitude, but users by about 20x. It's going to get worse

If they have 1600 satellites in orbit but are adding 40,000 more satellites (with more capability), then it’ll get better with 20x users, not worse. But anyway, the question was about satellite visibility, not bandwidth per user. The person I was replying to was claiming there will be fewer visible satellites as they’d be moving to different inclinations, which just isn’t feasible let alone likely.

> Even if they keep increasing the numbers, they will probably move those new satellites in a much wider net

The only way to make a wider net whilst simultaneously adding satellites would be to raise the altitude of the satellites, which I am fairly certain is not possible for the existing satellites.

Reply on news.ycombinator.com