Hacker News Reader

Comment by sseveran | Hacker News Reader

Indeed there is. ISPs don't run a non-oversubscribed network. If they did your bill would be much higher. The costs are in building cables, maintaining cables, buying and maintaining local pop network equipment, building and maintaining whatever backbone they run, installing and maintaining network equipment in network meeting rooms where they connect with transit providers, etc...

View on HN

I think the suggestion is more along the lines of you buy a set bandwidth with a certain contention ratio (or ratios) and then the ISP lets you contend for the bandwidth as you like.

Having a bandwidth cap is basically the ISP punishing you for contending too often.

That said, assuming the ISP wants to keep an average speed higher than the minimum their contention ratio would offer, people who download a lot would have a greater impact on that, and could therefore cause the ISP to have to invest in more hardware and back-haul.

The counter is that the ISP isn't upgrading their network anyway, they're just juicing the market by charging some people more.

I agree with what I think you are stating implicitly that bandwidth caps a blunt instrument. ISPs don't really have a lot of nobs to turn without setting off the network neutrality crowd. I wish there were more QoS nobs that could be turned to allow for lower latency and traffic prioritization as that would probably have some interesting applications.

Most ISPs are constantly upgrading parts of their network. Earlier this spring AT&T came through my complex and installed fiber in every apartment. Average broadband speeds have increased significantly over the last decade, and for that to have happened significant network upgrades would have to have taken place.

I don't think this directly impacts a cost per unit of data transferred, as the poster was wondering. I agree that maintaining the network and ensuring it can meet capacity is a cost that the ISP passes on to their customers, but I would point out that this is why people pay a monthly subscription to their ISP.

To my mind the question would be "if an ISP has network with a fixed capacity, do their monthly operating costs differ between a month where they consistently utilize 50% of that capacity compared to another month where they utilize 80%?"

I suspect the cost to the ISP is the same in either case: they have purchased the equipment and bandwidth to support the network and those costs are fixed, regardless of how much of that bandwidth they are actually using. If that is true, this idea that there is only so much data per month to go around and if you use more then your neighbor then you should be charged more money would be fundamentally dishonest.

Consumer ISPs are effectively bandwidth resellers: they buy in bulk at wholesale rates, split it up, and sell retail quantities to individuals. Wholesale bandwidth is generally billed based on the peak transfer rate during the billing period, and not a fixed price.(1) You also can’t assume a fixed network capacity- ISPs will buy enough equipment to maintain service to their customers instead of having brownouts. This capex cost is directly caused by higher aggregate usage from new and existing customers, so it makes sense to charge them by usage as well.

(1) This is overly simplistic; these are individually-negotiated B2B contracts, so details vary widely.

I would think that would mostly apply to DSL ISPs that are just using a telcos infrastructure. I suppose AT&T Fiber may be charged by AT&Ts backbone for traffic transiting their network or something like that. The major ISPs AT&T, Verizon, Comcast, etc... are building their own networks at least in part.

Yes, the truly big backbone ISPs aren’t generally paying for interconnection to the others. Instead, they have the capital expenses of maintaining and extending that backbone to handle growing internet traffic in general. If they don’t keep up with demand, they’ll lose customers and potentially their settlement free agreements with the other backbone providers.

Where I live there are no competitors for the one cable internet ISP (DSL is not available and satellite is n competitive). As such they have not need to worry about losing customers to _competitors_.

Based on this article, I suspect that ISPs like Comcast and Charter are not extending their backbone to handle growing internet traffic. It seems they are charging exorbitant fees to discourage increased usage instead (in this article, the subscriber canceled their Hulu account and lowered the resolution of video supplied by Netflix).

Those are both consumer ISPs, and thus not subject to the same market pressure as I described for Tier 1 ISPs (without naming them as such). Similarly, consumer divisions of Tier 1 ISPs are driven more by local market forces than anything else.

Having only one ISP locally is typically either caused by market size or local regulation. Comcast does offer a national backbone and sells transit, cdn and other similar services through a subsidiary called Comcast Technology Solutions. The backbone at a national level has definitely received upgrades and they are advertising multiple terabits of capacity. Of course upgrading the transit network does not help if your local link is saturated, which is a problem in some areas.

If there is truly demand for unmetered internet someone will step in and fill the void. Some ISPs also use total transfer as a way to discriminate between business internet and residential internet.

I don't understand how this answers the question "Is there any cost that an ISP company has related to the amount of data transferred?". This all sounds like maintenance and they'd have that if I use 1GB or 1TB.

If every customer can use 1TB, it is more likely that many customers will be using more bandwidth at the same time, therefore requiring more infrastructure to ensure their connections still provide their advertised bandwidth

I don't see how that would affect anything. If we assume that most people consume the most data in the evenings (streaming Netflix, etc), no matter the cap, you would see issues if bandwidth was an issue. If I'm downloading ISOs of Linux during the day at 100% everyday while people are at work, big deal, right? But, if everyone is streaming at the same time, if there was going to be an issue, it would happen if caps existed or not because at some point during the month you're going to have nearly all your customers with data to use.

The way I've seen this handled in the UK (a while ago, but I think some ISPs still do it) is to have caps at certain times (either by having a limit or throttling bulk P2P downloads), but then allow unlimited downloads at other times, especially overnight

> But, if everyone is streaming at the same time, if there was going to be an issue, it would happen if caps existed or not because at some point during the month you're going to have nearly all your customers with data to use.

It's a giant bucket of 'it depends'. If the CDN network serving the content has placed hardware within the ISPs network, it's likely a non-issue. This was a major issue before folks like Netflix started moving their CDNs closer to the customer. You might have a hard time finding the articles, but I recommend looking at the pissing match Level 3 and Verizon got into years back.

>it would happen if caps existed or not

I'm not convinced. You are essentially arguing that scarcity of a resource has no affect on consumption.

Reply on news.ycombinator.com