Starlink - Questions not in FAQ for existing users

So, maybe a bit of a long shot, but im seriously looking into Starlink for both personal and business use and interested to know what the expected latency is from one side of the planet to another.

Im sure others have other queries that aren’t covered by Starlink’s FAQ.

So without further ado…

  • can someone with Starlink in the USA please try and ping a few IPs and tell me response time? ns1.telstra.net (203.50.248.6), www.iinet.net.au (203.173.50.151)
  • traceroute to same addresses?

What I’m keen to know is whether or not the latency from one side of the planet to another is comparable with terrestrial service or how it compares.

Aware that local-to-local traffic may see worse than terrestrial pings, but I’m actually curious to see if it may even beat terrestrial service when going to the other side of the planet due to more direct path and radio/laser/whatever travelling through vacuum or close to it…

One of my use cases for example is remote users in different countries getting back to services in our HQ here in Australia… 2-way VSAT is garbage (550 ms assuming the earth station is local, if not add terrestrial carriage) and terrestrial internet is often 250-300ms plus anyway… from say Africa or central Europe to Australia.

But, low earth orbit satellite is MUCH MUCH closer to earth and potentially latency is much better due to the smaller distance involved.

Cheers!

Naively looking at PeeringDB for AS14593 shows pops in Seattle and LA.
(Perhaps it’s out of date or incomplete somehow).

https://satellitemap.space/ can also show you base stations.

So, if PeeringDB is to be believed currently all traffic from anywhere in the world goes through Seattle/LA, and how it gets there e.g. using satellite or terrestrial internet is also unclear.

My understanding of plans for the service is that traffic up from the user and then down to terrestrial internet as soon as possible (and vice-versa). Since terrestrial capacity is just so much more affordable.


You can look at telus bgp summary here: http://aurora.on.tac.net (starlink as number is 14593)

I’m very interested in if PeeringDB is to be trusted in this case, if the only two pops are on west coast us, starlink is pretty much poop for anyone else.

Basically if your server is an Australia and your user is in Australia currently your packets need to go through west coast US… they just need a lot more peers.

1 Like

Yeah as I understand it they’re installing local base stations second half of 2021, and it does route via the starlink mesh otherwise(?).

It may be the case that the routing situation will change over the next several months, you can pre-order for Australia now and if I’m not mistaken its due to start delivery on the east coast in the next couple of months…

Various places in Australia we’re seeing 700ms round trip via cellular, so if starlink can beat that it may still be a win even for local AU->AU traffic in some situations :smiley:

Unless you live somewhere in the middle of nowhere, Australia is actually relatively well connected when it comes to companies.

Google and various other hyperscalers have their own fiber, and for others, you can typically buy transit if you can reach ix.asn.au POPs.

Perth to Melbourne should be around 40ms another 10ms to Sydney.
Sydney to Singapore/Tokyo/LA ; should be around 130ms.
Perth to Singapore should be 50ms.

These are RTT times.

Amount of bandwidth over these links is theoretically pretty much unlimited - if you wanted a few hundred gigs, you could pay to have some network engineers plug in a few of your transceivers into DWDMs and run some config commands and get it tomorrow.
If you wanted a few terabits you might have to wait a few weeks because of global logistics and time needed to ship dwdm hardware.

My point is, 700ms RTT to anything within Australia is not due to physics, it’s just mismanagement, greed, price gouging and political lobbying - there’s no technical reason to go above 50ms within AU (unless you count money as technical).

Theoretically, lasers in space could get this latency down further by another 25-33-50% . Light speed is lower in fiber than in free space + laser only curves a little due to gravity, and fiber need to navigate seabeds and land access rights.

Latency from dishy to satellite should be around 2ms, and then it’s all about additional routing in space aka. “mesh” + ground stations + latency from ground stations beyond that.

So if starlink wants to offer lowest latency internet in/outside of Australia, they don’t really need to mesh one satellite to another, they just need to make sure they peer in a lot of places in Australia (having lots of IX ports in different cities helps) and they need to carry users traffic to those places themselves over high quality links, either their own, or leased.


Also they need a few more satellites to be able to provide reasonably non-intermittent service and to be able to serve more users. (they have about a 1000, and plans to go to around 10k before end of year).

I don’t know what the satellite-satellite throughput is - is it worth the hassle and energy to use it for low latency cross continental links. High throughput links need lots of power (signal to noise ratio) and I don’t know how much throughput you can get from their particular brand of blinking lasers over whatever distance they have between satellites today.

I’d love to know if customers get a public IP , and they can access the modem interface to change settings…

most certainly not. will be a CGNAT unless they hand you IPv6 as well.

EDIT: confirmed CGNAT + IPv6

damn… so unless ipv6… no chance to reach your self hosted stuff.

theres always tunneling a VPS.

2 Likes

From what I’ve read the satellite to satellite laser links have only been deployed on a small number of satellites operating in an almost polar orbit. They are necessary to serve areas on the poles that have no ground stations.

… or just straight up tunneling a-la tailscale.

Starlink has no coverage on poles at all, they’re very low earth, nearly polar (if you can call 60 degree near), sun synchronous orbits. But, not letting them talk to each other sounds like a wasted opportunity, worth checking if they explain that somewhere.

considering what it is, it’s quite impressive. So I think “unless IPV6” is a very small concession.

1 Like

My question is how it handles doppler shift. I know RF transmission systems for news helicopters have a hard time already with doppler shift and ATSC terrestrial broadcast hates doppler shift, basically being unviewable in motion. Wonder how Starlink handles this in case it’s in motion in a Motorhome or RV.

In fact, how would it handle doppler shift at the speed of a jetliner?

Same way you would at the speed of a car.

Modern technology doesn’t have an issue with it, it just has to account for it.

Eh, if it’s based off of common modulation it might have issues. Not very well versed in RF tech but I do know news helicopter coverage does depend on mitigating the doppler effect a lot if they’re using their own private airwaves and receiving equipment. Even wireless steadicams have this problem if it’s moving too fast.

1 Like

No coverage yet but they do have some satellites in a polar orbit.

More info:

It’s not about where I live - it’s about the remote mine sites in the middle of nowhere that we work at :wink:

And yes, they’re in the middle of Australia, mostly northern half, hundreds or thousands of km from anywhere.

For background - you’re dealing with someone who’s worked with remote areas networking in Australia for 20+ years…

Agreed. But it is what it is - the only connectivity in some areas we have presence is cellular through ONE telco who has the monopoly on comms at that site. Typically these are often microwave backhaul and haven’t been upgraded in many years and we do not have any sway over the telco to do so.

The reason the service is shit doesn’t matter, it is what it is (outside of our control) and starlink is an option that will come available shortly.

Until starlink our options have been 2-way sat (500ms, trash, expensive), cellular (not always available, but usually; oversubscribed, limited data caps, single monopoly telco has the rights to put a tower on the mining lease) or paying through the nose for fiber connections (again, limited telcos due to ownership of the only fiber run to the site - if there is one).

AS one example location:

Another:

(big dunes in the way from nearest mine to complicate microwave bridge… there is literally no infrastructure there yet, but we start work in a month or so).

We also do a bunch of work in places like Zambia, Mali, Kazakhstan, etc. Which are even worse. 2-way VSAT service there is 10 grand a month or so for 1-2 megabit with 950ms latency back to HQ.

1 Like

Cheers for that, so looks like we’ll have some other work to do.

Still, IPv4 over IPv6 tunnel is something we can work with. Additional work, but as above… options are limited and the work required is worth it if we can get sub 500ms and decent bandwidth. CGNAT is fine for generic internet, IPV4 over ipv6 IPSEC tunnel will be fine for private network (essentially, if the service is fast/low latency enough, I’ll MAKE it work via whatever hackery is required).

10+ megabit on fixed line is thousands per month in these areas (like… 100s or 1000s of km from nearest town - “just run a fiber” is an N million dollar project) - and that’s if it is even available. National parks, etc. are a thing.

@thro sounds like you need to test starlink out. From what I’ve read (see article I linked above) they don’t have widespread laserlink deployment and probably wont for 1.5 - 2 years so you’re experience will depend on the ground stations they have “near” your location(s), but it could be alot better than your current options.

1 Like

Yeah, we already have one on order, will certainly be testing. :slight_smile:

Just trying to get ahead of any gotchas

Sweet! Well let us know when you get setup how well it works!