43% with less than two choices for 25mbit is more concerning. 25mbit is around what you'd need for a family of four without being bandwidth crunched all the time.
I'm curious why you think 25 mbps is what you'd need for a family of four? Is there some sort of data out there for this?
Netflix states you need 3.5 mbps for SD, 5.0 mbps for HD.[0] With a family of four, does this means everyone could be watching HD Netflix, and do various other browsing simultaneously, and still be under 25mbps? If so, wouldn't 25mbps be far above what a family of four would need "without being bandwidth crunched all the time"? (Assuming that HD video would be the most data-intensive thing) Especially since, likely more often than not, everyone in the house will not be streaming HD video simultaneously.
The reason why I state this is because, while I get more speed is always better, we seem to often state X mbps is the bare minimum that should be achieved, but never say why. It makes it seem we are letting the perfect, or great, be the enemy of the good in some scenarios. For instance, if 10, 15, or 20 mbps is perfectly fine for most households, then shouldn't we use those numbers instead?
> Assuming that HD video would be the most data-intensive thing
I think this is the wrong assumption to make. I have a 30mbps-ish plan right now, and I can slow my apartment's connection down to a crawl by initiating a large download.
25mbps might be enough for a family of four whose connections are consistent, but I'd wager that the average home's traffic is more on the bursty side.
That could be solved with better QoS on the router right?
Obviously a download from a good server will saturate a 100mbps line, but so what? Other than video games and massive file backups, 25mbit is fairly fast. I pay for faster, but it's only a marginal improvement.
Better QoS on consumer routers would definitely help, but it's not a panacea -- ballooning website sizes also lead to bursty network conditions, and I wouldn't be surprised if my router's QoS is configured to prioritize HTTP(S) connections. Same goes for automatic updates.
Edit: I meant to add this:
We have the ability to provide (or at least develop) 100mbps connections to the vast majority of Americans, and we recognize that Internet transfers aren't getting any smaller. It's good to look at ways to improve the status quo (e.g., QoS), but we should also be looking to change the status quo itself.
Your edit is essentially the point of my original post. Yes, if we can get 100mbps for the vast majority of Americans, that would be awesome. But, if the vast majority of Americans would be fine right now with 10-15-20mbps, wouldn't it be better to analyze our national broadband capabilities based on that benchmark? From what I understand, 100mbps is overkill, so why use that as a benchmark?
I come at this with the thinking that, it's not "us" that's building out the infrastructure, but the telco companies. So, really what we are doing is analyzing the build-outs telcos are doing, and judging whether it is good enough or not. Is it really fair to say that, if there aren't 3+ telcos providing 100mbps, but there are 3+ providing 15mbps, that's not good enough?
Well, I disagree that the vast majority of Americans are fine with 10/15/20mbps. They might get by with it, but I'm sure they notice their streams buffer or drop in quality when they do (or their system does) something bursty.
In light of the obscene amount of money that taxpayers have sunk into ISPs, there's no need for that. They may be the ones building the infrastructure, but they're taking advantage of us along the way.
I ball parked it conservatively. I thought netflix streams took more like 7.5mbps. So I bet they could probably get away with less.
I bet a 4 person household will have less than 3 streams almost always. So that's 10mpbs. You'd need some room for surfing, updates, downloads, etc. I bet 15mpbs is fairly usable for a family of 4.
Not to be an old man, but back in my day, we had one PC that everyone shared. And it had a 33.6kbit modem that was UPGRADEABLE to 56k. But the upgrade was a lie. And you had to disconnect when your mom wanted to call her friend. And we liked it.
Yea, certainly wasn't trying to call you out or anything :)
And yes, I remember when there was only one or two TVs in the house. Then there was 56k, where only one person could be online, and it took up the phone line. That's not to say we shouldn't strive to greatness, but it's amazing how far we've come.
Which leads to another line of thinking: If a family does only have a 5mbps connection, and is thus forced to share their streaming, maybe by forcing them to all watch the same thing in a room together like we once watched TV, is this so much of a bad thing? Are there other ways to mitigate a low mbps so that you can still experience a 2018 Internet?
This makes me sound like a telco shill, but the 1gbps/100mbps/Xmbps-or-bust arguments, I think, are sometimes utopian in nature and often discount how much people can do without such high-powered Internet. Yes, of course more speed is good, but are those with only 15 or 10mbps really in the stone age?
However, I say this while living in Seattle, where we both have amazing choices and speed, so I am definitely one of the fortunate ones.
I changed from 50Mb/s to 1000Mb/s service a few months ago; latency is much the same, which means that when the pipe is not crowded, I really can't tell the difference.
(But it's wonderful headroom for large file transfers.)
I can't tell the difference between 75mbit and 900mbit. Most servers don't get anywhere near 900mbit. And my hardware is almost always a limiting facture (wifi, hdd speed, etc).
One way to do a test is to bittorrent something legal, say an Ubuntu DVD.
If you use a transmission client (e.g.), you can change the upload speed of your client. Do one test at a low upload speed (5K/second), and one test just above your ISP's advertised upload speed.
In the first case you should see your speed peak out around 850mb/s. In the second, if you have a bufferbloat situation, the download speed will drop significantly as the uploads push the download acks into a queue, forcing TCP to throttle heavily downwards.
It might actually, mostly due to buffer bloat on some modem stacks.
The anti-buffer bloat stuff from the recent DOCSIS (I forget if it was 3.0 or the one after) are supposed to help a LOT with latency issues. Not as good as fiber but close enough that global distance issues become dominant again.
The "buffer bloat" guy has been useful. I've been arguing against big FIFO queues for decades, but he's more into PR and visibility than I am. Any place you have a major choke point, especially home router uplink connections, you need QoS, or fair queuing, or something, or queue delays will go through the roof. At least send the ACKs first.
Latency can definitely improve from where we are at. The speed of light is different through different materials, that's why some people are using hollow fiber optic cables to have lower latency connections. Laying a more direct line would also improve latency, my connections to Seattle backtrack about a hundred miles east before heading out west again according to traces. Faster routing hardware would shave off a few milliseconds. Economic factors hold back latency improvements for the majority of internet users, not hard physical limits.
Speed of light isn't the only factor in latency. Quality of the equipment, routing protocols, etc are far more important since those are the things we can actually change.
I strongly suspect that for an average consumer, latency is driven by the TCP stack on every piece of hardware close to the edge (local machine, router, modem, ISP point of presence) to a much higher degree than the transmission medium or backhaul. Consumer internet has a long way to go before it becomes HFT backhaul-grade.
