There is an interesting unbalance because Comcast has so much leverage by owning the last mile, they can push around Tier 1 providers. I'd like to fix that, mostly by creating a public policy around municipally owned <i>Layer 1</i> infrastructure between customers in their cities and a city exchange building. Conceptually it would be no different than the city owning the sewers and outsourcing the water treatment plant to a contractor (or two). Creating a new "ISP" would involve installing equipment in the City Exchange(s), providing compatible customer premises equipment to subscribers, and then patching their 'port' at the City Exchange to the ISP's gear.<p>Its going to be a long conversation :-)
> Nobody paid anybody for the service because it was assumed to be symmetrical: as many bits were going in one direction as in the other so any transaction fees would be a wash.<p>The justification for peering is not equal traffic, it's equal value - my customer wants to communicate with your customer. Regardless of the direction of traffic, the traffic is equally valuable to both of us because the traffic is the primary thing our customers are paying us for.<p>Unless, of course, I can get you to pay me for it anyway because of some unrelated advantage such as the fact that your customers can leave you more easily than mine can leave me. Comcast and others are attempting to leverage exactly that - in many regions they have no viable competition whereas Netflix and L3 are much more replaceable in their respective markets. This is a prime example of abuse of a monopoly.
Would be very interesting to see what happened if the big CDN providers just depeered Comcast for 24 hours. Would certainly cost Comcast a fair amount in customer support calls, bad publicity, and properly bring the debate to the general public.
If a customer is paying for an internet connection, they are paying for access the full internet, to the best of their ISP's abilities. This is the net neutrality law we need: ISPs should be compelled to upgrade their backbone links as they become congested, to satisfy their customer's demand. Congestion can be easily monitored and often these peerings are "free". (Yes there is a non-zero cost to increase switch and router capacity and to have someone plug the cables in, but it's not like Level3 is charging for the bits exchanged.) But the point is, since most ISPs are de-facto monopolies in this country, we need rules telling them they have to upgrade their capacity to meet their customers demand, if they are promising broadband speeds.
This cable-menu style image from the comments is scary: <a href="http://i.huffpost.com/gen/1567010/original.jpg" rel="nofollow">http://i.huffpost.com/gen/1567010/original.jpg</a>
This wouldn't happen if those ISPs didn't have local monopolies. Networks should be opened by selling traffic wholesale to other companies so that they can compete for subscribers on those networks. The network owners would have more than enough money for upgrades and if they don't, downlevel ISPs will sue them.
Somebody has to pay the money to upgrade the equipment and bandwidth available at these exchange points. The very reasonable argument in this article is that the ISPs should pay that cost, which seems reasonable given that their customers are demanding it. It sounds like the ISPs are playing a game of chicken, trying to see if their peers like Level3 will throw money in to pay for the ISP to upgrade its equipment and bandwidth. That's certainly something the ISPs can try to do; on the other hand, what are their customers going to do, <i>not</i> use Netflix and YouTube? If a pile of customers of one ISP start reporting that they're all having a poor experience with high-bandwidth video, and there are a pile of well-publicized press releases blaming the ISP, customers will start complaining to the ISP, and they'll have to upgrade their infrastructure eventually. (And in areas where they have competition, there's an incentive to upgrade before the competitors, to avoid losing customers; while there isn't such competition in every locale, there are enough locales with more than one ISP choice to make those customers painful to lose.)<p>But what does any of that have to do with mandated peering requirements at the NSFnet exchanges? Who would enforce that, and why, when any two major networks can set up peering at any number of meet-me rooms? Requiring that an ISP peer as much traffic as is available or not peer any at all seems ridiculous; some ISPs will suck more than others, but that's the problem of them and their customers, not a problem for the entire Internet.<p>Meanwhile, I'm surprised there aren't more startups and VCs looking to bet that "new ISP that doesn't suck" is a viable business model. People are chomping at the bit for Google Fiber, which seems unlikely to grow to a national level without developing competitors. This is a space with very few competitors, and there hasn't been serious competition in that space since DSL stopped being a viable option.
You would think, okay, if Comcast is terrible at maintaining sufficient peering for it's customers needs -- and if the OP proposal to throw Comcast out of peering exchange points happened, that would certainly lead to increased terribleness for it's customers -- then eventually it's customers would choose a different ISP. The market would solve it.<p>The problem is that in many many markets, Comcast (or another ISP) are pretty much the only choice. Customers don't have another option, no matter how much Comcast underfunds it's peering infrastructure or gets thrown out of peering exchange points.<p>So what is the consequence to Comcat for underfunding? What is the consequence to Comcast for even such a disastrous outcome as getting kicked out of the peering exchange point? Not a lot.<p>I'm not sure what the solution is, but 'regulate them as a common carrier' is certainly part of it, since they are a monopoly, and the common carrier regulatory regime was invented for exactly such a monopoly.
I've previously interned at one of the mentioned Cable Companies, and I see both sides.<p>The solution is to make it 'capitalistic'. Change all of our internet contracts from Unlimited (up to 'x' GB/month), to a simple $/gb cost.<p>It would be in the ISPs best interest to provide their customers the fastest internet connection as possible. E.g, if a customer can stream a 4k video vs SD then the ISP would make more money per unit time.<p>Think of it this way, if comcast charges $0.25/GB, and a netflix SD show is say 1GB and HD is 4GB, then comcast
grosses $1 for HD and $.25 for SD for the same customer streaming request.<p>Over time its likely the price per GB would decrease, just like it has for cellular.<p>On a more evil side, this would also stop chord cutters. Pirating content is no longer 'free', and Netflix would cost significantly more than $10/month ($10/month + 'x'GB * $/GB).<p>As for what rates to expect, if comcast charges in ATL $30-55 for 300GB, that'd be about $.10/GB to $.20/GB. As for speed tiers in a $/gb system, your guess is as good as mine.
While I agree with the general thrust of the article, there is one fallacious argument here.<p>Cringely argues that cable breaks even and money is made on the net, but that's an artificial distinction. What if cable disappeared? Would they still make money if they had to pay for the upkeep of the network with only Internet fees? The desperation and risk of this game of chicken convinces me that the answer might be "not much." The loss of cable might very well be apocalyptic for these companies, at least from a shareholder value and quarterly growth point of view.<p>What's happening is very clear to me: the ISPs are trying to either harm the Internet to defend cable or collect tolls on streaming to attempt to replace cable revenue. That's because cable is dying a slow death. This is all about saving cable.<p>The fundamental problem is that cable ISPs have an economic conflict of interest. They are horse equipment vendors that got into the gas station business, but now the car is driving out the horse and their bread and butter is at stake.
The proposed solution is at the bottom of the article (which is why everyone seems to have ignored it):<p>> The solution to this problem is simple: peering at the original NSFnet exchange points should be forever free and if one participant starts to consistently clip data and doesn’t do anything about it, they should be thrown out of the exchange point.<p>I do have a couple questions though - who is in charge of the original NSFnet exchange points, and do they have this authority?
I don't know why everyone is up in arm over this. Here is reverse thinking and a perfect oppoturnity for everyone.<p>The current situation is that Comcast doesn't have the equipment/resources to handle extra internet traffic at its peers. Most people want Comcast to buy more stuff to handle it, why don't we think the opposite way- get Comcast to decrease its amount of traffic?<p>If we can get Comcast to consume less traffic, they wouldn't have to complain to other peers about load asymmetry.<p>The best way to decrease traffic? Make Comcast has less customers.<p>Why does Comcast has so many customers, even though their resources cannot handle it? Because they have a government mandated monopoly in the last mile, so they are forced to have more customers than what they can handle.<p>We can come to a conclusion that last-mile monopoly -> network congestion -> forcing L3 to pay for peer.<p>If Comcast has to compete with other ISPs for last miles, the traffic load would shift from 1 single entity (Comcast) to 10+ smaller ISPs. In such case the traffic load problem would not exist.<p>Another solution is to breakup Comcast.<p>See? This is a perfect opportunity. Comcast can has its multi-tier network, but at the price of the last mile monoploy. After all, if they want to have the right to choose peers, we customers should also have the right to choose ISPs.
The UK experience is, broadly speaking, that any ISP who is sufficiently tall to have the appropriate interconnects can offer a service to a customer via the incumbent's last-mile infrastructure. (This is for telephone-based ADSL broadband - the UK's only cable operator is not bound by this.) But, furthermore, competitor ISPs are enabled to install their own equipment and backhaul directly into local exchanges, known as LLU - local loop unbundling. LLU allows competitor companies to provide just your broadband, or your voice telephone service, or both.<p>There is one further step, whereby the prices of the incumbent monopoly are regulated in areas where no competition exists. Ironically this works in the opposite way to how you'd think, as it forces the incumbent NOT to offer their lowest prices in that market - the intention being to make monopoly areas prime targets for competition and to ensure that potential competitors aren't scared out of the area by predatory pricing.<p>It's an odd system with good and bad on both sides, but it seems a lot better than being stuck with a single source of Internet access.
People keep claiming that there should be a "free market" for bandwidth...but then they say that the ISPs should have to absorb the costs of peering (which can be significant -- the hardware isn't free) without passing the costs on to <i>anyone</i>. The backbone providers complain when the cost is passed to them; the consumers scream bloody murder when the costs are passed to them in the form of a bill.<p>Obviously, there's no free market in the status quo: we (consumers) basically expect to pay a low, ever-declining price for bandwidth, while someone else eats the costs of a growing network infrastructure. There's an economic disconnect, and legislating that it shouldn't exist seems worse than futile.<p>I say: pass the costs on to the consumer, and break down the monopolies on last-mile cable service. If the cable companies had to compete for subscribers, they could still pass on the costs of improving their infrastructure, but they'd have to compete with everyone else to do it.<p>In other words, the problem here isn't "net neutrality" -- it's that we've got a monopoly at the last mile that we need to destroy.
Godaddy, Rackspace, Google, Amazon, etc. have skin in this game. With multiple redundant network connections they could, for a day or a week, defend neutrality by shaping their traffic to the lowest common denominator or routing their traffic to avoid the peer's punitive bottlenecks. Today its Level 3 and Netflix, but tomorrow it could easily be them.
"and make their profit on the Internet because it costs so little to provide once the basic cable plant is built."<p>That's some big hand waving, because laying the cable costs a fortune, and takes many years to recoup the cost which is why there is so few are competing for this "super profitable" business.
If the peering ports are congested that means that either the ISP needs to add more ports, or they are oversubscribing their capacity. Just make it illegal to sell more capacity than you have and the problem is solved.
I like the article overall but I don't understand the author's proposed solution. The issue as it stands is apparently a lack of peering, in that big ISPs are using transit to reach large content providers rather than directly peering to those networks. So how would "kicking them out" for a maxed out connection work? If I buy transit from Level3 and my connection maxes out, I'm no longer allowed to be a customer of Level3?
The extreme download vs upload traffic asymmetry between Comcast and L3/Netflix has been mentioned several times as a straw man argument for why Comcast is justified in charging Netflix directly.<p>Maybe Netflix could find some creative uses for all that idle viewer upload capacity to reduce the deficit ;)<p>- Have every Netflix client cache and serve chunks of the most popular streams P2P-style. You could have a DHT algorithm for discovering chunks or have Netflix's own servers orchestrate peer discovery in a clever way, for example by only connecting Comcast customers to peers physically outside of Comcast's own network. This would reduce Netflix's downstream traffic and increase viewer uploads.<p>- Introduce the Netflix-Feline-Image-KeepAlive-Protocol, whereby every Netflix client on detecting a Comcast network uploads a 5MB PNG of a cat to Netflix's servers over and over again while you're watching a video. Strictly for connection quality control purposes of course.
The problem is not to peer or not to peer. The problem is WHERE to peer.<p>I work for a european ISP and the problem we have is the location of the peering. Big content providers will happilly peer with you in, say, Palo Alto or Miami; but they will refuse to add a peering connection in Europe. Why? because today the problem is about WHO pays the Intercontinental route (which limited and is expensive bandwidth).<p>Level3 is known in the industry as a pioneer for bit-mile-peering agreements. This means you have to sample the origin and destination of the IP packets and make some calculations to know how many miles the packet has traveled and pay / get paid if someone dumps long haul traffic to a peer. Getting to this is complicated with current tecnhology and many companies are refusing to peer with Level3 because they don't know what will happen with their business with bit-mile-peering agreements.
Along these lines, can someone ask whether net neutrality ever existed at all? Akamai and F5 have been helping big corporations like Disney circumvent internet bottlenecks for over a decade now. Those who have had the money have managed to purchase faster delivery schemes for over a decade.
Could it be, then, that telecommunications companies are consolidating so that they can extort money not from the small guys, but from the big guys? Are Hulu, Netflix and others willingly submitting to the extortion because they see no other way out?
To be sure, the telecommunications industry is in desperate need of regulation because providing good service at a reasonable price for a reasonable profit is not good enough for them.
Since everyone is pitching their own solutions, how about I post mine. Let's take the example of Netflix and Comcast. Instead of no deal with Comcast, and thus giving Comcast Netflix users really slow or no service, Netflix should make the deal for now, and tell subscribers that if you use Comcast the Netflix rental is higher. By passing off the higher costs to the users, Comcast customers are given the incentive to switch ISPs.<p>Everyone shows loss aversion, and so will be determined to find out why being on Comcast gets them penalised. They will learn about its dick moves, and complain to Comcast to make them remove these fees so they can access Netflix, which they have already paid for access to.
> It’s about money and American business, because this is a peculiarly American problem.<p>Hardly. We've experienced the whole interconnect brinkmanship locally too (South Africa). Its actually quite the opposite - the interconnect things are a lot nastier in other countries because it tends to be paid for (powerful co vs underdog) whilst the bigger US setups seem to run mostly open peering.
I can't help but wonder if the RICO Act applies to this sort of extortion. My first thought was FCPA, but none of the ISPs involved can likely be construed as "foreign officials." The behavior can be described as demanding kickbacks, however.
Just for scale, backbone links these days are not 10 gigabits/sec, more in the order of 40-100 gigabits/sec.<p>The Amsterdam Internet Exchange is the largest and most important exchange in Europe, and it's peak traffic each day reaches 3 terabits/sec.
Maybe I'm naïve beyond any recognition, but shouldn't the ISP's or whoever is peering charge based on the bandwidth amounts? It sounds like they have a flat-rate contract with each other and now they're charging more?
Level3 should drop the same percentage of outbound packets from Comcast, that Comcast drops on the inbound. If every tier 1 did, Comcast's internet service wouldn't look all that good any more, would it?
It will be interesting to see how gaming download services such as Xbox Live, PSN, Steam, etc would be affected as there as the file size of video games gets larger due to advances in the video game industry.
So on the one side is the fat-cat ISP who doesn't want to make expensive capital investments ih their transport.<p>And on the other side is the fat-cat vc funded video content providers, who don't want to pay for the their mp4-based saturation of all the pipes.<p>This is a negotiation. There are two active media campaigns that are trying to gin up our anger against The Other Guy (tm) as part of their negotiations. I just can't get invested in this nonsense.
1. Peering is based on equal traffic both ways.
At the moment we tend to download gigabytes with a few bytes of request. As video-communications really takes off (yes chicken and egg - see below) this will get lost in the noise<p>2. rise of ad-hoc local networks
This might come out of mobiles, this might be me dreaming, and it might come with sensible home router designs, but ultimately most of the traffic I care about probably originates within 2 miles of my house - my kids school, traffic news, friends etc<p>A local network based on video comms - that will never happen. just like mobile phones.<p>3. electricity and investment
In the end this is down to government investment. Let's not kid ourselves, gas, water, electricity, railroads, once they passed some threshold of nice to have into competitive disadvantage not to have, governments step in with either the cash or the big sticks.<p>Fibre to the home, massive investment in software engineering as a form of literacy, these are the keys to the infrastructure of the 21C and it's a must have for the big economies, and it's a force multiplier for the small.
<i>Except it’s actually right (not wrong) because those bits are only coming because customers of the ISPs — you and me, the folks who have already paid for every one of those bits — are the ones who want them.</i><p>What is the source of the notion that, because you paid for your consumer broadband, all bits are paid for and the charge for carrying them cannot be split with the other side of the connection? Why is it so bizarre that both sides of the connection have to pay for it? Because you're used to your phone working differently?<p>As an analogy, you know how you used to pay for a subscription to a magazine and there were ads inside which advertisers (the other side of the connection via the magazine in this case) also paid for? The magazine split its fee in two: you paid part of it, and the advertisers paid the other part. It's the same here.<p>There is nothing fundamentally wrong with charging both sides. You may prefer a different fee structure but a better argument than "I already paid for it!" is necessary.