Power users, businesses, researchers, and others might expect more, but not the majority of residential users. Muni fiber investments will last quite a while.
[1] https://www.marketplace.org/2019/08/27/for-faster-internet-d...
Comcast in my area doesn't even make claims of upload speed anymore, but my recollection of when they used to is that it was about 10% of the download speed. If you want more than 2 Mbps upload, that 20 Mbps plan wouldn't cut it.
There is a full duplex version of DOCSIS, but its adoption faces some challenges due to how operators have to re-architect their physical networks to support it.
There is also currently a push to extend the downstream spectrum up to higher frequencies, enabling both more downstream bandwidth and also the ability to move the upstream split higher, therefore increasing upstream bandwidth.
(Disclaimer: I work with a client who develops products in this space.)
I think I'm around 50 Mbps and it could be used much more heavily, but it doesn't make any sense on YouTube's end to send me a 50 megabit video stream. It's a the storage and upload costs on the provider side limiting that.
Even if you pay extra for Netflix's 4K support, that only needs 25 megabits. It's compressed to hell compared to a 4K bluray, but if we all had faster internet I don't know if that would change. It's lot harder to market the actual visual quality of a video compared to talking about how many pixels it has.
Maybe we'll see whole new services that need the bandwidth, but most stuff that sends a lot of data could be downloaded in advance if you don't mind waiting 5 minutes instead of having it instantly accessible. How many billion dollars are we willing to spend to build out faster networks to avoid that? What's the selling point for very fast streams, we go back to cloud mainframes and all of our devices are dumb terminals that stream our games and the whole OS from an AWS datacenter?
Of course the actual pricing, availability, etc are also important but I don't think there needs to be a one-size-fits-all model for Internet bandwidth.
Even if the physical plant supported symmetric bandwidth, I would expect different pricing plans for different utilization patterns.
And I can't even get a faster more pricy connection because my legacy contract gives me upload speeds I don't get till I add 50% to my bill
- data costs for bandwidth costs hasn't gone down at the same rate
- A big chunk of an ISP is labor costs, which do not go down over time
It depends. If you're constantly doing new customer setup/teardown, yeah that'll cost man-hours. But once your network is up and running it really doesn't take much work to keep it running.
My ISP is a very small local outfit. They have a "few thousand" customers... and the whole thing is run by literally two dudes, doing everything from netops to hw installs to billing and support. Yeah I'll have a few hour outage once every few months, but overall the service is great and it really doesn't take that much effort to keep things ticking.
Basically rent a quarter cab in a colo, get a router (or two), a switch (or two), and a server (or two). Negotiate with last mile provider and a pair of upstreams. Get your DC to x-connect you to your upstreams and last mile provider. Do some marketing (elevator ads in residential apt buildings are apparently basically gold). Buy a pallet of cable modems, spend a few weeks configuring your crap... $50k later, you have an ISP, and you can offer service at 50% of the big boys' rates while still raking in a substantial profit.
For clarification, at the business level data has always been charged by capacity and not volume. And the costs have fallen at double digit rates since they've been measured.
http://drpeering.net/white-papers/Internet-Transit-Pricing-H...
ISPs plan their networks for peek capacity, the only reason they charge by volume is because they can.
Explicitly the reason for automating things, esp. since network connections don't require a ton of labor once provisioned.
Orchestration engines help to do the former (Ansible, Salt, Puppet, whatever), and they usually tie into monitoring systems like Nagios, ScienceLogic, or SolarWinds, and can trip, and then launch, remediation efforts automatically.
For instance, in terms of $/GB, data storage costs have been strictly decreasing in nominal terms even with the existence of inflation.
That said, I am not sure how much innovation is going into the costs of maintaining a fiber network, so it’s possible that it won’t decrease as fast as 2-3% a year on average. I would be really interested if someone had those numbers!