Item — HN Reader

I previously worked in engineering at ISPs. There are two points to make: (1) Monthly data caps do not help manage congestion, at least not directly. However, non-engineers have a lot of trouble understanding this. (2) Telling customer support agents to say "this is not about congestion management" is a perception-management move. Cable companies are very sensitive to criticism that they are offering a "shared pipe" and that you're somehow not getting what you signed up for.

On the first point: Congestion is problem during peak hours, not during the whole month. You could download a few terabytes over the course of a month and not cause a problem for anyone else, if you avoided peak hours. Data caps only work for congestion management if they discourage you from using services during peak hours.

There's not a grand conspiracy by the cable or mobile companies about this, though - it's mostly a failure to understand this concept, or an unwillingness to go back to peak load pricing (e.g. different night/weekend rates like phone plans have/had).

6 comments

Satilite ISPs actually do this. The one I had when I lived with my parents gave us 18 GB a month with unlimited downloads between midnight and 5:00 am. Since I ran Linux I'd just throw my heavy downloads in crontab and go to bed and have my big shiny package the next morning.

Interesting! I did a search and found one example - Exede? It's encouraging to see some experiments with this.

The need to exempt various types of traffic for business reasons is ultimately why I think monthly data caps will fade away for cable/telco-provided Internet service.

An example would be some provider like Disney saying that in order for Comcast (or some smaller operator) to be able to carry their TV channels they must also zero-rate streaming from their sites. Or Hulu and Verizon do a deal in which Hulu content doesn't count toward a cap on Verizon phones. These things are becoming more common, and I think the difficulty of figuring out which bits count and which ones don't (especially when they all come from the same CDNs) will prove to be more trouble than it's worth.

Exede seems to be moving away from this model, actually. Their new plans don't offer unlimited periods at all, but instead, some types of traffic are exempt from the caps. The problem with that is Exede won't say exactly what type of data is exempt other than "basic web browsing and email." No one really knows if that includes HTTPS, HTTP2, QUIC, WebSockets, etc.

My home town's ISP (http://pris.ca) does this for its wireless customers. Transfer between midnight and 7AM doesn't count towards your 10GB/mo cap.

Lots of Australian ISPs use this model. For example: TPG Broadband Off-Net 50GB -> 10GB peak / 40GB off peak (2am-8am)

10Gb would last me less than a week of browsing.

It lasted less than 2 days when I had a plan like that. Only could change to 100gb since that's all was offered. Now I still go over constantly without trying. The ISP gets congested massively on Friday, Saturday & Sunday even with these limits.

peak load pricing? in an age where physics and manufacturing processes have given us native 100GE signals at shockingly low cost per bit?

it used to be backhauling access traffic was expensive (think T1s to cell phone towers). well, those days are long behind us. a little less profit taking and a little more infrastructure investment would solve bandwidth problems in duopolist markets.

It's not the backhaul that is the chokepoint. It's local nodes near your house. The one you share with 50-100 homes or more.

A Netflix "ultra HD" stream requires 25Mbps. Even if all 100 homes are simultaneously streaming ultra HD, that's 2.5Gbps in total. Coax can handle that easily with modern equipment.

Obviously they could be using very old equipment that can't, but whose fault is that?

Cable companies can't replace network hardware every three years. It's just not profitable. Cable systems were upgraded to broadband 2000-2005. 10-15 year replacement is short in the telecom industry. It's not surprising that new entrants into the market can give much higher speed. If you demand upgrades every 5 years that is going to increase costs considerably.

Also a large amount of coax spectrum is for video tv service.

The real curiousity is why Verizon fios can't deliver near 1gbps. It's already has a fiberoptic network. I'm guessing it's the hardware at the nodes.

Quite so. Usually the same in mobile networks, but sometimes backhaul is a limitation there still.

It's getting a bit ridiculous. ISPs have pocketed multiple rounds of order of magnitude jumps in network technology while customer speeds have stalled and "usage caps" are on and off.

And that's just the "assume they are not acting in bad faith" stance; we know they are willingly not upgrading the interconnects to (eventually) Netflix and others, who their customers are requesting more and more data from.

>or an unwillingness to go back to peak load pricing

It's actually a decent system - for eg. one local mobile operators provides unmetered/uncapped 4G between 22-6 for a small premium - if you don't have access to a landline it's a great deal.

So why don't they do something like: deprioritize traffic in proportion to how much you that user uses at peak times. (Shouldn't be hard to turn it into a more rigorous formula.)

So the pipe is normally shared equally between requests, but a heavy user might get only half a normal share. That way, they optimize against the real scarcity (in peak time usage, not total downloaded) but no one has to worry about overage charges, and any throttling is to the benefit of other users.

There might as well be a conspiracy to advertise best case performance and avoid making any sort of guarantees about minimum bandwidth (for consumer plans).

I understand that this is because the one is more marketable than the other, but that isn't really very comforting.

In India, Internet from Asianet and BSNL used to be free between 2am and 8am for some plans