Updated: Jan 3
Has There Ever Been True Net Neutrality?
Even though I closely followed the “neutrality” debates over the past few years, I only recently realized
how concrete its impacts could be (in my case my kids crying because they could not watch their favorite movie on netflix). The screaming of my youngest son begging for Peppa Pig to start again, along with the recent scrapping of the so called “net neutrality” by the FCC provided me with a new viewpoint on the debate that has been shaking the video streaming industry for years: Did net neutrality die out 10 years ago, when the first SVOD services popped out on the market? Let us look back in time for a while. To make the above statement clear, one needs to make out how the different actors interact with each
Beginning of ages
At the beginning of ages (ie in the 00s), life was simple. You would subscribe to an Access Network “A”, and your neighbor to an Access Network “B” (an Access Network, a.k.a. an Eyeball Network, is an Internet Service Provider selling Internet access to end users: Cable Operators, Telcos). In order for you to reach the machine of your neighbor, or to reach “content” (such as emails, newsgroups, etc.), every “network” had to pay a B2B ISP. This ISP would also give access to the rest of the world: Internet. Internet was expensive, bit rates were low, content was poor, yet the ecosystem was simple.
Peering between Access Networks
With network technologies improvements (QAM/Docsis, xDSL, fiber, …) the bandwidth and the number of subscribers grew significantly. Access Networks (and especially cable companies in the US) began to investigate solutions for connecting their networks together, by concluding what we call a “peering” agreement. Peering is a mutual “free deal” where one Access Network will open up all the routes of its own customers to another Access Network, and vice-versa. In other words, your machine will “know” how to access your neighbor’s equipment, without going through an Internet Service Provider (Level3, Cogent, Tata Communications, ...)
When concluded, these kind of agreements are “free”: neither party will pay its counterpart in exchange for traffic. This is known as the “sender keeps all” philosophy, where each Access Networks makes money by charging only its own customers. Those peering agreements started as “private peering”, meaning that both Access Networks will connect to a switch, between their entities, usually hosted in one of the Access Network premises. As the number of Access Networks began to explode, the latter started to connect themselves into neutral locations (where the costs would typically be shared) that could host many ports. These locations are known as “IXP” (Internet Exchange Points, ~500 sites worldwide), that can be seen as the synapse of the Internet.
Of course, peering is based on the fact that the impact on the network is “balanced”. There might be some discrepancies (because Access Network A has more subscribers than Access Network B, for instance), but usually the peering agreement remains valid until one Access Network generates significantly more traffic than the other (a 2.5 ratio is usually considered as the upper limit). Moreover, costs for peering are usually shared: if the peering capacity needs to be increased, new ports/switches are added and the upgrade cost is shared between tenants. Last, but not least, free peering and transit are not incompatible: ISPs and Access Networks usually have a peering agreement (so that the ISP can reach the Access Network subscribers), while the Access Networks purchases transit from the ISP to access the Internet.
CDN & Peering with Content Providers
Similarly, content providers began to generate a significant amount of data. They started to investigate peering agreements that, at first, were beneficial to everyone:
Content providers could reduce transit costs
Access Networks could give their end users privileged access to the content: lower latency, higher bandwidth (as the traffic would flow from the content provider to their own network without going through the ISP)
For the sake of simplicity, and although their business model is pretty different, lets’s merge CDN providers (Content Delivery Networks: Akamai, Limelight, …) and pure content providers in the same category for this section. CDN started to emerge by leveraging the fact that the end user usually pulls similar content from the internet. They provide an answer by caching the data as close as possible to the users. As this traffic is generated by the content providers, it makes sense to merge those actors (in addition, some content providers have developed their own CDN …)
For the above reasons, these peering agreements were seen as beneficial for both tenants, and followed the same rules of being “free”. This peering is probably the first scratch to the “net neutrality” model.
Net neutrality: a not-so-clear definition
According to Wikipedia, “Net neutrality is the principle that governments should mandate Internet service providers to treat all data on the Internet the same, and not discriminate or charge differently by user, content, website, platform, application, type of attached equipment, or method of communication.”
More specifically, it means that the ISPs should guarantee that they don’t favor or disfavor access to the content based on its provider. As one can guess from the previous section, the fact that some content providers have peering agreements with the Access Networks implicitly means that its subscribers have a privileged access to some contents … compared to other contents whose providers do not have a peering agreement. As long as the volume of content remained low, this was not an issue … until video came in.
Video Streaming and Paid Peering
Youtube kicked off video streaming on the Internet in 2006. It’s been a major shift in networks usage, as video objects are of course much heavier than usual web content. Hence, CDN gained a considerable traction, as caching data became mandatory to deliver on-demand and live video to end users. CDNs are now one of the most significant players in the peering ecosystem. Let me quote a few figures from the Cisco VNI to emphasize that statement:
IP video was 73% of all IP traffic in 2016 (expected 84% in 2021)
52% of all Internet traffic have crossed CDNs in 2016 (expected 77% in 2021)
Back in 2006, Access Networks, and especially cable companies, felt that the free peering model was short of breath. Keep in mind that initially, those peering agreements were made to provide reciprocal access to customers from different access networks, without going through a 3rd party ISP. But the benefits have to be reciprocal as well, and the network traffic has to be balanced for the free peering model to work. However, internet video is of course massively asymmetric. When you pull a video from yousomething.com, the traffic basically flows only in one direction (from the CDN to the access network).
Large cable companies started to meter that traffic and started “offering” paid peering deals for the “partners” that did not meet its free peering agreements. Akamai and Limelight had to subscribe to those peering deals in the late 00s to be able to deliver the content to Comcast customers. Another famous power struggle is the Comcast/Netflix story that is hereinafter summarized:
Step 1: Netflix was using CDN services to deliver video. Associated paid peering costs were growing at the same rate as Netflix customers: exponential
Step 2: Level3, acting both as a CDN and ISP, won the Netflix distribution deal. As Level3 also have peering agreements with ComCast, and as their trafic suddenly grew by a significant ratio (because of Netflix), they request Comcast that some additional peeing ports are added.
Step 3: Comcast didn’t accept to increase the peering capacity. Indeed, the traffic they were used to receiving money from (through paid peering with Akamai) was now flowing through a free-peering agreement they have with Level3.
Step 4: Months of public war and millions of complaints of Netflix users who could barely watch their content through Comcast network. This was an uncomfortable situation for both actors, but Comcast knew they had one key advantage: the captivity of their customers.
Conclusion: Level3 agrees to pay for peering. Overnight, Comcast customers could watch House of Cards in HD, and this opened the way for other access networks to conclude paid peering agreements with Netflix.
Back to Net Neutrality
At a first glance, paid peering is not in contradiction with Net Neutrality: there is no segregation per se; The access network will not disfavor access to any content. That being said, some specific content will be delivered with a higher quality of service because the access network receives direct money from these content providers. This also allows access networks the ability to ask a higher fee to their subscribers because they guarantee a smooth access to the content, which is clearly not compatible with the Net Neutrality principle.
Note: I just terminated my xDSL contract with my provider and moved to an ISP that has peering agreements with Netflix … Just to discover that the ISP I had just left had also reached a peering agreement with Netflix the next day, too :)
Quortex changes the paradigm by introducing "Just In Time Everything", a technology that allows you to control your CDN costs by leveraging compute resources in your infrastructure.
Interested ? Contact us!