Television has gone broadband, and that transition will continue over the next decade as more content finds its way onto IP networks as opposed to the old-school dedicated pay TV networks. As this transition has unfolded, new players like Netflix, YouTube and Hulu have entered the market while both ISPs and big content companies are trying to figure out how to protect their revenue and adapt to this new era.
Amid the confusion and competition comes an Organization for Economic Cooperation and Development report full of data that digs into the peering debate that is currently causing customers of certain ISPs pain. This is an issue that, like an earworm, keeps popping up and drives anyone looking to consume internet delivered content on televisions nuts.
As I laid out on Wednesday, the issue concerns how internet giants interconnect. Right now some ISPs look at interconnection points as a source of potential revenue — a way to get a company sending large amounts of traffic through to the ISPs end users to help offset the cost of maintaining and building the ISP network. Thus, they want to implement a paid peering model for companies that send a lot of traffic their way.
Paid peering is not a new concept, but it is relatively unusual, according to the OECD. In 2012, the group released a survey that showed that 99.5 percent of 142,000 interconnection points it studied peered with each other for free. The new study discusses the economics of peering in a way that’s hard to come by in on-the-record conversations with people who manage peering relationships. And it presents a compelling argument against the common rationale offered for paid peering by ISPs.
From the report:
To carry 500 Mbit/s in one direction, a 655 Mbit/s or 1 Gbit/s connection is needed irrespective of whether the traffic in the opposite direction is 10, 100 or 499 Mbit/s. The same equipment would be needed in a network to carry traffic; as for the core of that network there is no asymmetric networking equipment, supporting higher speeds in one direction. Accordingly, there would be no difference in costs. A further consideration is that it is unclear how ISPs would make their traffic balanced, with most ISPs currently only supporting asymmetric up and download levels of, for example, 20:1 on ADSL2 networks. An alternative response from a content provider could be to strictly limit the traffic it sends to the ISP, such that it is always in balance. The result could be that traffic would dwindle to close to zero, because most of the traffic from an ISP consists of control messages controlling the flow of incoming data. When there is less incoming data there is also less outgoing data.
In short, the OECD is saying that idea that web traffic can ever be balanced between two networks is a fallacy given that the end users consume way more than they create and broadband speeds are generally asymmetrical to reflect that reality. Plus it asserts, implementing balanced traffic would be a catch-22 because much of the outgoing traffic from an ISP is to manage the flow of incoming traffic, so as incoming traffic drops so would the outgoing traffic the application is trying to balance against.
But before breaking down some of the economics and the question of balance, the report also shines a light on the question of motive — suggesting that it has changed over time.
It points out that a few years ago most of the peering arguments involved getting smaller traffic-generating sites to pay to interconnect, with the argument that it costs money to interconnect to smaller networks. In many ways the 2012 study was a rebuttal to that argument, showing that more network connections generated via settlement-free peering benefitted both smaller and larger providers (although the smaller providers did get the lion’s share of the benefit) as well as the internet at large by expanding its reach and influence while lowering costs for all participants. From the report:
Another justification used to not peer between two ASs, or that a network is flooded, is that the traffic is unbalanced. In many peering relationships and especially with connected television services, one network is likely to send more than it receives. What is unclear in this case is why the traffic from, for example, the content provider to the ISP has to be balanced. Historically, the case was made that networks that could not reach a 2:1 or 3:1 ratio were too small to peer with larger Tier 1 networks. With the advent of large content providers the argument has reversed, now end-user ISPs argue that networks exceeding a 2.5:1 ratio should pay for peering.
One way to get around paid peering is to use a Content Delivery Network or transit provider that already peers with an ISP. In the last three years content companies have tried to move away from that model as a way to cut their costs and get better control over the network. Google, Netflix and even Facebook are building out caching servers, trying to build their own CDN services that will carry their content to the “front door” of an ISP. Many ISPs welcome adding a server from one of these content companies as a way of cutting down their own transit costs for getting the content on their network. But some do not.
ISPs argue that they don’t want to support boxes from “every” provider of web content, and say that the boxes take up space and consume power. And that’s the tension that can hurt customers, as ISPs demand paid peering and content companies demand that ISPs use their content servers. Consumer pain has become a negotiating chip for internet giants to use in an attempt to squeeze in a bit more revenue or lower costs.
The report also covers data caps, threats to the connected TV hardware business and rebuts the idea that over the top video will destroy broadband networks. It’s worth a read.