Bit Traffic: Who Pays for Internet Traffic and Why? | Hacker News Reader

Bit Traffic: Who Pays for Internet Traffic and Why?

thepbxblog.com

30 pointsjosh260037 comments

TL;DR - internet traffic is paid for by users, but high-bandwidth use like video costs the ISPs more, and they'd like to double-dip by also charging sites like Netflix.

Yes.

It's important to note just how much video content has grown compared to revenues though. If you click through to the link on the Deck, you'll see a bunch of stats about video utilization compared to revenue. The numbers are pretty mindblowing.

Costs for consumers haven't really gone up, but utilization has increased considerably. Any idea about what the answer should look like? (Note: you can't say Google Fiber).

>> Costs for consumers haven't really gone up

Costs for consumers are whatever the ISPs decide they are. The answer should look like "if you cost us more, we charge you more," just like every other business on the planet.

It doesn't have to be X bits = Y dollars. It could factor in load; the meter runs faster at high-demand times and slower at off-peak times. That solves the "bandwidth is a renewable resource" argument and the "some people use more than their share" argument. They'd just have to ensure that the meter is highly visible to consumers.

Let's be frank, the only real cost for consumer internet is the last mile. Since Internet access for residential bandwidth was reclassified thanks to the memo from George W. Bush around 2004 we haven't had real competition because common carriage doesn't extend to your front door.

Contrast this with Datacenter bandwidth where paying $.10 a megabit isn't unheard of. The cost structures are not related to the core infrastructure, but rather the edge.

I think this problem is a lot more complex than you're making it out to be, but that's just my opinion. In a lot of cases the rates for ISPs aren't set randomly, they're anchored to standards set by regulatory bodies (particularly in the case of DSL).

In short, pricing bandwidth is hard. The real cost is the last mile.

I think you understand this better than I do, so maybe I'm being simplistic. But isn't the gist of what you're saying "it's complicated because ISPs have gotten themselves a near-monopoly status and then, in reaction, we've regulated them?"

If so, that doesn't make me sympathize with them at all. If they want freedom to charge more according to access, or to charge more according to the difficulty of running the last mile connection, I'd say "give up your monopoly protections and we'll remove your regulation."

If they don't want that, I'd say "you start charging both the consumer and the content provider and we'll pass a regulation saying you can't."

If a company wants the government to be their sugar daddy, they shouldn't complain when they get grounded. You want your own rules, you move out of the house and compete in the free market.

In very old industries you often have players who have a monopolistic status. Markets either become price efficient or competition deficient over time. In the case of telecom there is stratified warfare.

At the consumer level, competition is scarce. In the SMB Segment competition is fierce. In the Datacenters world competition is fierce, but at the core of everything you have to go through 721 Folsom street to get out of the left coast.

This fight is particularly interesting because it's one of the first to cross the Commercial-Residential boundary. On the one hand, consumers pay for access, on the other, large ISPs peer networks. One is a contract, the other a handshake.

In this particular case, ISPs argue that since usage increases so should rates, but in order to arrive at such a conclusion, the large ISPs would have to ignore their monopoly.

Let's take this a step further though; there's a really big problem with everyone building their own network. That is, no one wants thousands of wires running to each residence, but that's what a competitive last mile would look like.

The answer is municipally controlled networks with private operators. The infrastructure would be owned by the state, but operated by competitive virtual operators.

Unfortunately, this vision is impossible in America, and that's why we can't tackle this problem straight-on and instead have to operate through layers of abstraction.

TL;DR: public networks, private operators is ideal. Impossible in US.

Why do you say it's impossible in the US?

It's impossible due to regulatory and lobbying pressure. I believe Palo Alto, CA operates a network similar to this, but that kind of network isn't possible in a city like San Francisco or LA.

IMHO the incumbent carriers have no interest in competing if they can be a monopoly.

If revenues are not rising, then they need to raise prices and/or alter their business models to fit usage.

I kinda expect unlimited data plans to be a very short lived phenomena with the possible exception that it may end up free where the traffic can be offloaded to a wifi.

Others have mentioned that the answer should look like "charging more to those who use more".

But I see it differently the relevant scarcity is over bandwidth: total data transfer at any moment in time. That is, downloading when everyone else is "hurts" the network a lot more than when you download when it's below capacity.

So why not just throttle heavy users when at full capacity, and only then?

I know a lot of people here don't like anything involving throttling, but it only seems fair that (if you don't want metered usage), then people who put less load on the network should get priority when there's only a scarce amount to go around. (It should be weighted, of course, by how much of your usage was on or close to peak hours so that you're punished less if you schedule your torrenting for off-peak.)

What am I missing?

It's not scarce. Power isn't scarce, copper isn't scarce.

Access is scarce. This is not a scarcity problem, this is a distribution problem. Bandwidth inside of a Datacenter can be had for $.10 per mbit, but because access to your home is prohibitively expensive, you pay more.

Does that help? The key point is that it's a distribution problem, not scarcity.

I think there's a confusion of terms here: Whatever is a "problem", therein lies the scarcity; you just need to identify it. If there were no scarcity, everyone could have what they wanted.

So what is it that the ISP cannot provide, specifically? I'm guessing it's the sum of all data passing to/from its customers per unit time. Hence why I called bandwidth the relevant scarcity, and suggested apportioning it based on how much users are using the network.

Can you operationalize the term "distribution problem" so that it's clear what specifically the problem is? "It's a distribution problem" is easy to fall back on, but hard to make precise. They say the same of food shortages, for example, but people have different ideas of what it means.

I don't think it's the Sum total of all data.

I think the problem is that the last mile access is prohibitively expensive. Glass connections in a datacenter are $.10 per megabit, but AT&T DSL over copper to your house is $30 for 1.5 megabits. That copper could become glass, but no one will pay for that upgrade.

The distribution problem is that in order to compete with at&t at the residential level, you have to build a parallel connection and the cost of doing so is prohibitively large. Therefore, everyone is forced to rent access from AT&T, BUT since common carriage doesn't apply to residential access lines, AT&T doesn't have to share.

If sonic.net could build glass to your house, they would, and they wouldn't ever complain of congestion, but because of artificial barriers to entry no new competitors can enter the market.

Does that make sense? I'm not trying to be evasive, it's just a complicated problem. Carriers don't have a problem dealing with aggregate capacity, they have a problem dealing with edge access capacity. Core carrier routers like the CRS-3 from Cisco can move 322tbps+, and they do, but you can't put a CRS-3 in every DSLAM.

It doesn't make sense, because I asked specific questions, and you seem to focus more on showing off your vocabulary than answering them. Let me rephrase:

Forget for a moment about technology/capital upgrades, and assume that the ISP wants to give everyone as much speed as possible. The customers collectively try to download as much as they can, as fast as they can. Naturally, they'll hit a limit. At what point does the limit affect their usage, and how can the ISP fiddle with access so as to discourage overuse? And why is my solution not one possible method (if suboptimal)?

:( sorry.

Let me try again.

When we talk about wireless bandwidth, there is a big problem with congestion. When we talk about residential bandwidth, congestion isn't the issue so much as competitive access is. It's not that AT&T and Comcast have the best access, they just got there first and are able to block other competitors.

Throttling will make the ISPs more money, but it doesn't help consumers or solve this problem.

I think this might be explicitly what you're looking for: At the pricing and access tiers available today it is virtually impossible to saturate and block a local connection (at the residential level) such that you impact other subscribers at a regional level.

Throttling doesn't solve this problem it just reduces usage. ISPs have the overall goal of reducing usage because it increases profit. The logical fallacy is that while your model proposes an ISP that would want to give everyone as much as speed as possible, the reality is that all ISPs want users to use as little bandwidth as possible.

Does that help? Really I'm not trying to show off my vocabulary it's just a complex subject :(.

Well, sort of. Depends on what country you are in, too.

In AU/NZ at least (and I am simplifying) you have the last mile, the trunks between exchanges, the handover points between big exchanges and ISP POPs, and finally you have the ISP upstreams - any of which might be a bottleneck if sufficient traffic is generated.

In AU/NZ (and much of the rest of the world) you have local peering exchanges etc, but ISPs also have this cost for "upstream", which is money they have to pay to their friendly international cable monopoly to get bits from faraway places.

These backhaul links are not normally configured to handle all users running at full rate. That would be expensive, and silly.

In the fixed line world, you might observe "congestion", but it the root cause is usually "oversubscription".

Everyone oversubscribes backhaul and upstream, the question is how much you can get away with. The reward for guessing lower than your competitors is lower costs (or being able to defer an upgrade cycle). The penalty for guessing way too low is churn.

I say "guess" because even though telcos tend to have good models of user behaviour, user behaviour can change very quickly (from a telco's POV) when new services come on line.

Such as online video...

Do the "costs for consumers" take into account the $300 billion in public money that was supposed to be put towards improving broadband in US?

http://www.newnetworks.com/broadbandscandals.htm

Now my own opinion: double-dipping like this is a stupid, stupid concept.

If an ISP were to charge Netflix for consuming its bandwidth, it would be like a taxi service slapping fees on an airport for generating so many rides.

"DARN those selfish airports, they're driving my taxi business under with their huge demand for rides!"

If ISPs can't figure out how to make more money from providing more service, they should hand in their business licenses.

Yes it is dumb, but another way to look at your analogy would be if there were only one road to the airport and that road was controlled by AT&T or Comcast.

Imagine that for a second and consider what impact that might have on your ability to hail a taxi :D.

All kidding aside, you make a fair point, but I don't think the playing field is quite as level as you're making it out to be. You can't just start an ISP that can peer at the same level of AT&T, you know?

The taxi/airport analogy is almost there but needs one extra tweak for correctness. In that analogy the ISPs would be the state / local government that built the roads. They would be trying to extract additional money out of the increased traffic, because one the roads are built it's harder to set up toll booths because of stringent regulation.

The issue is, and this going to be a huge tug of war for a long time to come, is the role of ISPs; either as 'content providers' or as merely a utility provider. One provides huge profits, one is a low margin business.

In the real world, tolls on highways are run typically by the local governments, or in the case of around where I live (nyc) 'public benefit corporations'.

Publicly traded corporations, obviously, don't want to be turned into government run utilities, even though this would probably be in everyone's best interest eventually.

Well then the answer is clear, the market has spoken: they are utilities. That's how they were setup, that's the bargain they made with We the People, and quite frankly, they don't offer content anyone wants (they aren't Netflix).

"Ooh, I want pay per view and Mediacom/Cox/Comcast digital movies", said no one ever. If they want to provide content, they should not be allowed to anti-competitively discriminate against other content providers. This is exactly what Network Neutrality is about.

But how do you force a publicly traded company to become a utility when it's in their best interests not to?

The stick is regulation, and the carrot is money.

Unfortunately, it's really hard to do once the govt. money is spent. I'm familiar with AU/NZ. Both are pumping significant public money into broadband infrastructure, and they took different approaches to enforcing fair network access.

In AU, they created a new company (NBN co) to run the "utility" part. See: http://whirlpool.net.au/wiki/nbn

The government is paying the monopoly incumbent to switch users onto the new network.

In NZ, our privately owned largest telco (Telecom NZ) semi-voluntarily split itself into two parts - a retail and a wholesale branch. The shareholders agreed to do this, because that meant the wholesale part (Chorus) would get a big govt. contract to lay fiber to the door.

In theory, the split is "arms length", so other network providers can access the infrastructure on fair terms.

Reference: http://www.stuff.co.nz/business/industries/5858284/Telecom-s...

It remains to be seen whether this is the right way to go about things, but both approaches do seem plausible.

The ISP's received lots of money from the government to build the "roads", except they didn't. They didn't invest in the required infrastructure and instead created monopolies and provided poor service and are now complaining they dont have the infrastructure. So now you have a country who invented internet and should be at the forefront of every aspect of it, lagging behind a lot of other countries in terms of infrastructure.

I take exception to your implication that the US lags behind other countries in terms of Infrastructure.

There are two classes of Internet access in the US: Residential/SMB and Enterprise.

We have a second world Residential/SMB bandwidth economy, but I would argue that the US has the only true first world Enterprise Bandwidth Economy. If you think of laid capacity in terms of terabits, the US is the clear winner.

The ISPs built massive highways but they didn't build the community roads necessary for normal people to tap into all this capacity. The bottlenecks for residential bandwidth are at the DSLAMs at the bottom of your street and not in the core routing systems for the biggest carriers.

It's tragic, but the US has a super stratified system. Enterprise users get amazing cheap bandwidth and everyone else is second class.

Operators in all countries are worried about reduced profits due to expensive data traffic, and all operators have paid large amount of money to governments for their licenses. So yes, they did build the roads.

Which would actually double-charge sites like Netflix, since those sites already have to pay for bandwidth/peering on their end.

This article is misleading. The "originator" refers to the sender of any traffic in any direction. Google and Netflix already pay an enormous amount of money to push their traffic to users. They don't freeload on the backs of local providers as those lobbying would have you believe.

They're just upset that as local providers they don't get money directly from Google. Instead they get a lot of money from their user base because their user base wants to get to Google.

To be accurate, Google doesn't pay directly for most of their own bandwidth. They have lots of their own fiber and peering agreements with most providers. Because of that, providers don't have to pay a backbone to send bits to Google, and Google doens't have to pay a backbone to send bits to the providers.

Of course, Google pays indirectly -- they have to buy, build & maintain those backbones.

This much is true, but the only reason a provider would to that is to reduce costs in the first place. It's their call to provide peering.

So I've never really understood why landlines should charge per data transferred. I mean, after all the communication lines have been run and hardware has been setup, the only additional fees are for power, right?

Lets say I'm an ISP and I have a gigabit connection to an internet backbone. I pay $X/month for it regardless of how much data is transferred. If I run fiber out to each of my customers that can run at gigabit speeds, how should I regulate/charge each user?

1. I could artificially rate limit them to less than the physical link speed. So I could let each user connect at 10 mbit/s. That would let me have many customers using there max speed at once and if I have 100 or less customers then they are guaranteed to get their full 10mbit/s. If I have more then they only get 10mbit/s at off peak times.

2. I could let everyone connect at the full gigabit speeds and put a quota on how much data they can send in order to prevent hogging.

Would you rather than 10mbit/s with unlimited data or 10mbit/s * 1 month of data transfer at a max rate of 1 gbit/s?

I imagine that you might want the first or second option depending on your use case.

[edit: forgot third option]

3. I could do fancy traffic shaping and make sure everyone gets at least 1/(number of active users) bandwidth and can burst up to full gigabit speeds when the connection is idle.

This seems like the best way to share the resource fairly.

See this is only with an even playing field. You're thinking about how you'd design a network in a world that didn't have all the embedded rules and dogma that we have in the US.

The biggest question is not network management policies, but last mile access. There are tons of ISPs with great bandwidth practices. Guess what? None of them have enough footprint to compete with the big guys.

The problem is not network management, it is edge distribution.

That powerpoint deck is awful. So video is an increasing percentage of internet bandwidth - so what? Price/performance of bandwidth is improving exponentially. It's unsurprising that non-video bandwidth isn't increasing as fast as video, since video is automatically big (and easy to make bigger: just move up to higher resolution) whereas "normal" traffic only gets bigger when users create more stuff. This increasing percentage doesn't say anything about increasing backbone costs, even though the graphs are meant to imply that it does.

That deck was done by a research group and yes I think it is intended to imply that the network costs increase as volume over the network increases.

While this is implicitly true, I agree that there isn't an explicit relationship, BUT it was interesting how dramatic the growth of video has been and how much of the network is video.

You do agree that carriers are moving more data than they ever have before, a fact which repeats itself daily. I'm not necessarily arguing that ISPs should be able to charge more for video, only that there is more usage on there network which may or may not impact their backbone costs.

You are correct that the implied linkage is misleading, but I wouldn't call the deck "awful". It's just a viewpoint, and it does provide some useful context/information.

the only way for isps to grow is not to block the progress but to embrace it, they should try to find a way to improve the technology behind their networks imho, could 4g networks bring something to consider here?

The problem here is Shannon's law. Ever go to a big event like SXSW and try to make a call? That's what massive deployments of 4G utilization look like; similar to downtown SF post-iPhone.

Wireless has a long way to go but fixed-wave might help here. There are also solutions from Ruckus and Meraki/Cisco which could help, but it's a big problem with no clear solution.

Reply on news.ycombinator.com