Starlink adds another 100Gbps connection to the SeattleIX
seattleix.net
seattleix.net
Combined with relatively low one-time port fees, the SIX's funding model has worked for 25 years - including dozens of large companies and bean counters: https://www.seattleix.net/contributors
I went to that page, see the top item is Starlink, the next link is "Welcome Danga Interactive AS32150". That's kind of a small AS compared to Starlink's 146K, but then I realize that used to be my AS, tummy.com.
At this point, most of the free up space is picked up by the big boys and you gotta do business with them.
Airn has a few senarios where you can pick up ipv4 addresses but it's pretty though at this point and you usually would need to be an established business with an existing ipv4 and a dual stack ipv6 setup. The ipv4 they are giving out is basically preferred to providers that are dual stacking and need additional ipv4 space to roll out ipv6.
Honestly, if you were to start today, there's basically no chance unless you gett off your wallet and pay a 3rd party at auction. And those ups are usually going to be trash. You'll spend a year getting them off blacklists and able to function.
So theoretically, you could get some but in practice, you're best bet is to rent them from Someone that already has them and put a router or server in their Colo.
An asn is not something you buy from a third party unless acquiring a whole corporation that already has one, since there is no real shortage of them, you'd get your asn from your regional registry.
Expect about $50 to 55 per IP now.
So 254 * 50 or about USD 12k to 13k? That’s the only cost and you get to own 254 IP addresses? Are there ongoing costs?
CAN'T BE HUSHED HERE
CAN'T BE HUSHED HERE
CAN'T BE HUSHED HERE CAN'T BE HUSHED HERE
But I agree that Elon time is absurd and irritating and sure feels like a scam. (I would argue FSD being sold for SIX GODDAMN YEARS on Teslas is _absolutely_ a scam.)
https://www.starlink.com/legal/terms-of-service-preorder#:~:....
Or are they happy to save their place in line?
It really started coming together in the fall and I would say since November it has been very usable. I would say it's easily as stable if not more stable than the 10/.7M dsl it replaces.
The issues I have with connectivity are probably 99% trees.
When I first started it seemed to need a total view of the sky to get consistent. As the software/flotilla have improved vastly, I use it for zoom and once every couple hours of zooming I may risk a 10 second blip.
I think they are getting better software wise of detecting the usable area of visible sky and utilizing satellites better.
I would think more satellites, more tree tolerance. I currently have my dishy too close to the road since I get the best view that way, I might try moving it back to the backyard this spring to see if it can work well there without worrying about teenagers using it for batting practice or some disgruntled cable employee or it getting stolen.
If you have an unobstructed view of the sky, it should be functionally as reliable as cable internet.
You can only pay the 500 for the dish once it's available for shipping.
And my relatives are not wealthy people -- this was half of their income for a particular month so at this point they'll be damned if they lose their place in line. They have no other choice -- not even DSL options in their area and the other satellite services have been so expensive and unreliable. They are completely at his mercy, and I feel like it's my fault for suggesting it.
You're defining "happy" as "not quite pissed off enough to leave". That's exactly the kind of game which (when compounded) makes fed-up people call things scams.
No I'm not, I literally mean happy. Starlink is by far the best ROI internet option for many.
You're taking a plausibly-most-likely scenario and using it to cast an assumption about the emotional state of hundreds of thousands of people. Don't you think you might be talking out of your ass?
It's possible that the delay is due to lack of legal clearance in the customer country. Starlink needs to get approval to install and operate the ground station satellite links in each country. From the few authoritative statements I've seen those are sometimes delayed due to local country bureaucracy.
I ordered Starlink in the UK in February 2020 and the equipment finally arrived in May 2020. I /think/ the stated delivery estimate at the time was 2-3 months and never reduced until the order confirmation email arrived and I paid the balance, after which the equipment arrived a couple of weeks later.
For example, Starlink claimed in early 2021 to be launching in Viet Nam "within a year", and have been selling pre-orders, but they can't legally operate in VN unless they form a JV with a local partner, since foreign companies aren't allowed to provide telecoms services there. The relevant government agency went on record saying that Starlink hasn't even applied for permission, yet they're still selling pre-orders.
In your Viet Nam scenario and similar cases I wonder if it is (partially) a result of the Starlink plan to use direct satellite to satellite LASER links to relay data in low earth orbit to a more remote existing ground station?
As in, if I recall correctly, SpaceX originally expected to be launching 'block 2' satellites with LASER links in mid 2021 but suffered internal hidden delays (that led to Musk's internal SpaceX existential warning in October 2021 and firing of some senior staff) so the first batch were only recently launched in early 2022.
Those LASER links being in place are likely a prerequisite for serving territories where SpaceX cannot obtain favourable permissions for ground stations.
Or people populate them at will
it's meant to be used by network admins at other ASNs that want to set up peering sessions across IXes and PNIs.
All Starlink customer traffic egresses via Google's network, and all their ground stations are along Google fiber paths.
actually most of their earth stations are colocated near long haul dwdm regen huts from older, more traditional big telco players like what is now lumen (former qwest/centurylink), level3, zayo, etc. The fact that they can buy transport circuits from those huts to the nearest major cities to meet google is useful.
by no means does google really own/run/control most of the long haul dark fiber and dwdm networks in north america. or even a good portion of it. the predecessor entities that have now been rolled up into lumen or zayo do.
Starlink is buying transport circuits from the traditional telcos who operate on the route to get to a major city with a Google POP.
Like, if you go out to the Prosser Regen hut site and starlink earth station, that's all legacy level3, now lumen.
You get much better prices if Google can see you already have most of your own infrastructure set up and ready to migrate to with not much effort.
That’s not true. Customers that were getting public IPs without CGNAT were coming from a non-Google ASN.
[1] https://cloud.google.com/network-connectivity/docs/router/ho...
My two Starlink terminals respectfully disagree. It has been mentioned numerous times on the Starlink forums. It's IP space previously used by Google Station.
it really depends on where you are and whether your terminal has been selected by (some unknown criteria) for spacex's network topology changes.
https://www.youtube.com/watch?v=N2ChhZcH2ok (slides: https://archive.nanog.org/sites/default/files/4_Reimer_Seatt...)
There's a number of places that are not out of capacity and if you put in the "right" street address it'll take you to the immediate full price payment order page.
the problem with terrible oversubscription on geostationary consumer grade vsat services is based on the fact that raw transponder kHz on a geostationary satellite is extremely expensive.
starlink appear to be taking a very different perspective since they absolutely know about the abysmal reputation of things like viasat/hughesnet/wildblue/etc in the consumer-facing ISP space, and won't ever let it go that way. because it would be a brand name killer.
If one fully understands the economics of geostationary it helps to better understand the tech behind starlink. I don't rule out the possibility that it goes bankrupt and ends up in a situation like the 2nd corporate incarnation of iridium, but the tech works great.
The fact that they just released a more expensive terminal to me means that they are trying to tackle the enterprise market where the cost per bit is significantly higher than with consumers, and those users will for sure get priority over consumers. It also means they are losing money too fast and they needed to at least have a solution that will potentially have higher returns in the short-term.
I agree the tech works great for what it is, but I also don't think it's been out anywhere near long enough for people to make judgements about how well it works. I don't think we will have enough longevity data on the terminal for at least a couple more years, so it's not clear if performance will degrade or if the motor will seize due to whether.
I think the users that got in now are having a great honeymoon period, but I don't think this will last, and we should not be comparing it to existing, mature services.
Maybe it’s related to your location. I’m in Vancouver.
Musk seems to have figured out that he can collect deposits for way more people than he could possibly deliver to, leveraging it as free operating cash. The cybertruck is a great example - millions of people gave him money for something that hasn't left prototype stage and will never be produced because it couldn't possibly meet federal vehicle standards in the US or Europe.
Starlink seems to be more hype too like his HyperLoop.
Even the video description is incorrect.
> Note the driver blows the horn as he enters the intersection. He obviously cannot stop the car. Probable FSD or Autopilot crash.
FSD/Autopilot does not stay on or prevent human interaction. Slamming on the brake turns it off, even assuming was on. This is a "unintended press of the accelerator pedal" type crash. Human error, and actually a good argument for why removing the human from the loop would be a good idea.
> Starlink seems to be more hype too like his HyperLoop.
Starlink is serving over 100,000 customers with good service all paying $100/month. Hardly hype.
I'm not aware of Tesla specifics, but i've previously read several stories on HN of car computers who wouldn't let the driver be in control. As a computer person who has moderate understanding of mechanics/electronics, i believe unless you have a physical kill switch to take control back from the computer, there is literally no way software can be safe enough that even in a crash you can regain control.
I mean apparently you've been around for a while on these forums. You're surely educated that software can never be trusted, especially with life-endangering situations.
That can't happen. People misremember things when they're in a panicked state. Or they're exaggerating to defend themselves. The instant you even slightly tug on the wheel autopilot and autosteer of any kind will disengage and revert to a traffic aware cruise control state (with an audible chime). And any application of the brake disengages autopilot/autosteer and also automatic cruise control with an audible chime.
> I mean apparently you've been around for a while on these forums. You're surely educated that software can never be trusted, especially with life-endangering situations.
I'm not sure what you're saying. People aren't "trusting" it yet which is why it's easy to disengage and override.
Yes yes i've heard that one before. I've had many bugs "that can't happen" over the course of my lifetime. Fortunately none of them involved a car so i'm still around to talk about it. A few years ago, somebody would have laughed at the idea of Boeing producing fallible equipment, especially since we as a species have decades of experience of air travel. Now would you still say it's unthinkable that Boeing could produce faulty hardware/software?
A link that gets posted on HN quite often contains many such bugs "can't happen" bugs, that aren't even caused by AI : https://beza1e1.tuxen.de/lore/index.html
I also remember some threads about high-range medical equipment (in actual hospitals) killing people due to software bugs.
> And any application of the brake disengages autopilot/autosteer and also automatic cruise control with an audible chime.
Is that a physical kill switch that triggers a sensor to produce an alert/chime? Or does the sensor ask politely the computer who actually controls the car to disable the autopilot? In the latter case, i know of no way to make this safe: for example what happens if the sensors requesting human control back die, short-circuit, or otherwise malfunctions? Or if the program enters certain kinds of memory violations? Or any other software/hardware fault?
> I'm not sure what you're saying.
I'm saying i've seen enough bad engineering (not always, but sometimes along with bad faith) across all fields i've been even remotely involved in. And i'm saying i don't even remotely trust the >100 micro-controllers you find in a modern car whose schematics and source-code we can't inspect. I mean i don't fully trust mechanical hardware either, but at least with it symptoms and failure modes can be easily reproduced and debugged.
And i'm definitely saying i would never ever trust a machine-learning algorithm with life-making decisions. More on this topic:
- James Mickens at Usenix on why AI is usually a terrible idea: https://www.youtube.com/watch?v=ajGX7odA87k
- jwz's blog (another HN favorite) is also full of examples of AI failures ; a fun harmless example: https://www.jwz.org/blog/2021/06/using-ritual-magic-to-trap-...
Additionally, we see just how a real human reacts to a real present danger: confusion. After identifying the threat the pedestrian instinctively runs into the path of the vehicle, just to run.
Why not 4x100G?
>Welcome Apple Inc. (AS714) at 2x1x100G.
Why not 2x100G?
>Akamai Technologies, Inc. (AS20940) has added 3x100G bringing their capacity to a total of 3x100G and 2x100G
And why not 5x100G.
How does one read these numbers?
From a logical construct perspective each "2x100G" interface is represented by a single "200Gbps" layer 3 interface. When referring to 2x2x100G, it really is meaning 2 layer 3 interfaces, consisting each of 2x100G ethernet interfaces in the link aggregation group.
Hashing of L3/L4 header data provides for consistent flow -> lag member mapping to avoid OOOs and other historical multi-link issues.
It gets better as you with faster connections and more users. At 100Mbps you can roughly do 20:1 over subscription, bump that to 1Gbps and 50:1 generally works fine. StarLink is aiming for roughly 50-100Mbps connections so this likely represents around 20,000 to 40,000 customers per 100Gbps link.
“SpaceX - Starlink (AS14593) has added 2x100G bringing their capacity to a total of 2x2x100G. They are connected to the route servers.” In other words they hit 40,000 to 80,000 customers and are preparing for another 40,000 to 80,000.
Meaning that ISP B will need to upgrade its ix port at a much later date relative to subscriber count, but its PNIs and cross connects are opaque and hard to know about from a public/external third party perspective.
One station feeds about one state on average, and I doubt they're feeding everything through that one connection.
That said there are some very significant content peers present on the SIX: Akamai, Amazon, Apple, Edgecast, Facebook, Google, Microsoft, Valve, among others. Hurricane Electric is also there and has an open peering policy despite being a very large ISP with a significant percentage of Internet routes and traffic.
I got curious and checked the local IX (smaller EU country), one of the largest ISPs has 4x100G connections, but their combined maximum ingress+egress almost never goes beyond 100G.
Dedicated gigabit bandwidth is still in the thousands of euros monthly, or in the hundreds in optimal situation (if you're already in a well-connected colo).
At a certain size you don't want 1x100G in case one link breaks, so for a large company they get 2x100G with each fibre going to a separate router. If one router goes down (planned or unplanned) you don't go offline. So now they probably added a second redundant pair.
You can see SIX's gear here:
* https://www.seattleix.net/topology
They have two Arista 7500-series chassis switches, one 8U and the other 12U (amongst other things):
* https://www.arista.com/assets/data/pdf/Datasheets/7500RDataS...
One of the line card options is 7500R-36CQ, which can handle thirty-six 100G ports.
That's not even Arista's highest capacity switch; the 7816R3 can handle 576x400G:
* https://www.arista.com/en/products/platforms
(I'm sure other vendors have similar product offerings.)
On the SIX side the ports are probably already in a redundant chassis but having 2 ports probably still scores you connections to two fully independent chassis in different failure zones.
2x100G to an IX are usually connected as separate ports, each with their own assigned IP ont he IX subnet. BGP multipath can enable load balancing across both ports. Otherwise you might have some peers configured on one port and others on the second. These days I like to have all peers on both ports and use BGP multipath.
If both ports are connected to a single switch on the IX side and the same router on the customer side then LACP can be used to combine them into a single 200G port with one IP. This is a less common configuration unless you are doing more than two ports (4x100G as 2x 2x100G lags for example).
There is another consideration: what happens if there is an icmp or arp flood on the IX? It's not supposed to happen but sometimes it does, even on the most major and well-run exchanges. In that case you might want to have all of your IX ports on one of your routers, keeping the other one "clean", just in case your control plane policing doesn't mitigate it.
Different networks place different priority on IX ports. In most cases it's ok if it's down for a short time (traffic will just shift to paid transit) so redundancy is not essential. In other cases it may be too much traffic to reroute so you care more about IX reliability and redundancy.
Been a lot time since I have done all this stuff though. I sort of miss it sometimes (though not the pager that comes with the territory).
Strictly speaking, IXPs do not work at OSI Layer 3, but usually at Layer 2.
All the Layer 3 routing logic happens in the routers of the members/customers of the IXP. The IXP is just there to provide a neutral interconnect to everyone without any filtering or policy.
The IXP runs router reflectors where people can choose to advertise themselves, and then other folks see who is in the neighbourhood and make their own routing policy choices.
The folks at Network Startup Resource Center (NSRC) have a series of videos on "IXP Design and Implementation":
They're by Philip Smith, who helped set up LINX (London, UK) in the early days:
Confusingly they list each of those as 200G capacity, not sure if that's a typo or it really is 2x 2x100G.
Not really sure why this is a big deal - IX bandwidth is very very cheap. It looks like seattleIX charges $7500 as a one off payment to add a 100gbit link: https://www.seattleix.net/pr20191218).
I would describe it as "capacity", not bandwidth. That fee is just to physically plug into the router. You have to make separate agreements with the other organizations also plugged into that router for peering or paid transit.
As an ISP, you will pay a transit company to transport your traffic to areas/operators you don't have connection to (very expensive, especially for small ISPs who get charged >1€/Mbit/s). So there's an incentive to establish direct connection with other operators your traffic may need to reach: that's what an IXP is for. Typically, you pay a membership fee to pay for services (electricity, network cabling), and then you can exchange traffic with everyone there for free (peering).
So there's really nothing new, unexpected or exciting about this announcement. It's just an ISP getting ready to exchange more traffic with other ISPs.
(Do even Y Combinator companies get as much coverage on HN as Musk's?)