Comcast has decided to start charging Netflix extra to connect Netflix’s customers on Comcast’s network. More or less. It gets complicated, depending on whether Netflix is being charged for data transfer, or interconnectivity.
The headline in the New York Times reads: “Comcast and Netflix Reach Deal On Service.” But Netflix CEO Reed Hastings posted on the official Netflix blog that there was a need for “a strong net neutrality,” calling the Comcast deal an “Internet toll.” That does not sound to me like Hastings came out of the deal happy.
Now, to be clear, what the deal is actually doing, on a technical level, is allowing Netflix to deliver its content directly to Comcast’s servers, rather than going through a middleman such as Cogent. It’s a type of “paid peering,” instead of “paid prioritization.”
Hastings, however, believes the two are the same thing – charging the content provider to provide the data at the rate that the ISP charges its customers. After all, the only reason that Netflix makes up 30% of peak Internet traffic is because that’s what end users request. Netflix is what Internet customers choose to spend the bandwidth they’ve already paid for, on.
What’s interesting is that while Netflix is paying for interconnectivity, other interconnectors to Comcast’s network do not. This is partially because of a “gentleman’s agreement.” Since most ISPs send and receive the same amount of data, there’s no reason to track who owes who what. But that suddenly becomes a problem when you have large interconnectors other than ISPs connected – the same data which traveled for free on the ISP is now charged a toll.
Hastings tried to make a point of the absurdity of the agreement that ISPs have by pointing out that if Netflix went to a peer-to-peer distribution model, it would upload and download at close to the same rate. Which seems to me less of an argument for not charging Netflix but more of an argument to keep track of all ISP traffic and charge all of them the same rate.
Doing so would require visibility into network peering routing and inbound and outbound traffic – to see what traffic is coming from whom and where. And this kind of monitoring isn’t just for managing costs but also to optimize the efficiency and performance of both inter- and intra-network transfers. For example, with this information, network providers can determine if switching some traffic to another network peer might get the traffic to its destination faster based on geography. Or they may discover that they are receiving way more traffic from a peer than they are sending which could prompt them to either balance the traffic or negotiate a different cost structure.
No one would have been able to predict the sheer size of Netflix. It’s clear we’re going to have to rethink the way we consider the real costs of networking going forward. Both sides have valid arguments: Netflix feels it is being discriminated against based on the sheer volume of its video traffic, whereas Comcast (and other network providers) face increasing margin pressure as they invest in building and maintaining the infrastructure required to transport it. And of course, consumers want some form of guarantee that they will receive services fairly, from both their Internet provider and content provider(s), whether they be established players or innovators.