Fibre versus Starlink Discussion

[scaesare]

Didn't get a chance to respond to the multi-casting thing earlier, but that was something we leveraged in the company I worked for that was a Satellite VSP (vertical services provider). This was in addition to the standard ISP internet services provided (with all of the latency-reducing tricks needed to get even mediocre performance).

We used multicast for provide content to endpoints for things like video loops (think store displays, business waiting rooms, educational content, etc...). Basically we'd multicast a video stream across the bird for each customer set. At the endpoint location we provided a cache/playback machine. A typical example was body repair shop: the cash register/POS system used the unicast satellite service for their credit card transactions, the waiting room would play an advertising video loop it had received via multicast the night before.

Multicast requires all of the devices in chain to support it, not just the transmission and end-points. This makes it viable for something like sat service, where you don't have other people's routers and switches in your path. It's a very efficient way to deliver a lot of static data to many endpoints simultaneously.

However that also constrains your set of services. As mentioned above, the nature of web surfing and on-demand media consumption is such that it's hard to do much with that via multicast. It would useful for "live" event channels where everyone is receiving the same event without staggered start times (akin to terrestrial broadcast), and then the endpoint can cache locally for pause/rewind, etc... It also is great for updating stuff: send the same firmware to 5 or 5 million devices so they can update. I could see that being useful for Starlink at some point.

Incidentally, the system we used actually shoved the packet payloads in to what looked like an MPEG DVB frame at the transmit side, as that what the systems in use had originally been designed to handle... the VSAT terminal on the endpoint then unpacked that in to a TCP/IP packet to had to the endpoint device. A hack, unlike the special-purpose transport used by SpaceX that @Cosmacelf refers to, but another example where the protocol traversing the bird is something different than just plain TCP/IP.

 

[Cheburashka]

Multicast in the sense of private internet is pretty pointless.

Starlink will not be faster or better than fiber, where it's available, because of physics and aggregate bandwidth limitations, but it wasn't designed this way in the first place.

 

[SageBrush]

Multicast in the sense of private internet is pretty pointless.

Starlink will not be faster or better than fiber, where it's available, because of physics and aggregate bandwidth limitations, but it wasn't designed this way in the first place.

I think you grossly underestimate the predictable behavior of the human animal. As one trivial example, although I do not watch TV series, I'm willing to bet that those who watch two episode of a series are highly likely to watch every episode. Storage and AI will make multicasting a valuable tool to amplify Sat bandwidth.

 

[scaesare]

I think you grossly underestimate the predictable behavior of the human animal. As one trivial example, although I do not watch TV series, I'm willing to bet that those who watch two episode of a series are highly likely to watch every episode. Storage and AI will make multicasting a valuable tool to amplify Sat bandwidth.

However that only makes sense with uniform local storage and an aware playback mechanism... something like Dish or DirectTV's DVR is an example. They can pre-cache the rest of a popular series to a separate volume on the hard drive in their box via multicast, and then you can play it at will locally via their same player built in to the box. That closed ecosystem is what was described in an earlier post for how they handled some pay-per view movies.

But for something like Starlink, the local router just hands data off to your computer... your computer doesn't know what to do with the cached data. What's more, when you log in to a multitude of different services/apps (NetFlix, Amazon, Hulu, Disney+, etc...)to play content they would have to be aware of the local cache rather than connecting to their respective motherships.

A Starlink-provided client you load on your computer could address some of the local storage issue... but only for that devices. The ability to plug a USB storage device in to the Starlink router might be a better solution for the local cache. But then there's still the issue of all the client software that would need to be modified to be on-prem cache aware. The router could try and proxy requests to all the services and intercept the content requests to try and deliver them locally, but that's fraught with a multitude of problems...

In practice, it's only likely to be successful with a content service that partners with Starlink to become multicast system-aware...

 

[bxr140]

As one trivial example, although I do not watch TV series, I'm willing to bet that those who watch two episode of a series are highly likely to watch every episode. Storage and AI will make multicasting a valuable tool to amplify Sat bandwidth.

Consider that example could be significantly better served by simply leveraging off-peak hours to asymmetrically exchange user/app-initiated data with users. There's plenty of surplus supply from the constellation to a given user cell over the course of a day. Its the peak supply that's always going to be the limiting factor--again, ultimately limited by PFD--and the multi-cast concept really doesn't work at peak. Maybe something like the World Cup or Super Bowl...but we're really talking way-high 𝝈 scenarios.

Certainly some theoretical multi-cast could optimize the otherwise under-utilized off-peak supply, but there's really no material benefit to do so for either the network or the users. Much like the electrical grid or really any kind of shared infrastructure, its all about peak.

Further exploring the human side of it, not everybody is getting into Tiger King at the same time, some folks are binging and others have lives, etc.; Each practical human variable makes the multi-cast solution simultaneously more complex and less efficient. Layer in the fact that any satellite service will only excel [over terrestrial services] in cells with a pretty low user density and The Network ends up with a massive number of small and likely wide data set to develop what need to be statistically relevant mutli-cast solutions.

 

[mspohr]

Ars Technica: AT&T lobbies against nationwide fiber, says 10Mbps uploads are good enough. AT&T lobbies against nationwide fiber, says 10Mbps uploads are good enough

AT&T is lobbying against proposals to subsidize fiber-to-the-home deployment across the US, arguing that rural people don't need fiber and should be satisfied with Internet service that provides only 10Mbps upload speeds.

 

[ItsNotAboutTheMoney]

Ars Technica: AT&T lobbies against nationwide fiber, says 10Mbps uploads are good enough. AT&T lobbies against nationwide fiber, says 10Mbps uploads are good enough

AT&T is lobbying against proposals to subsidize fiber-to-the-home deployment across the US, arguing that rural people don't need fiber and should be satisfied with Internet service that provides only 10Mbps upload speeds.

The problem isn't fiber v ADSL. The problem is the subsidy. If you must subsidize, subsidize wireless instead.