If you are reading this blog, you probably know what I mean by hot potatoes. The routing in the Internet is often referred to as hot potato routing. This analogy comes from how BGP and IGP work together so that incoming IP packets to an Autonomous System (AS) are treated like hot potatoes. If someone hands you a hot potato, what do you do? You pass it to the next person as soon as possible. This is what BGP and IGP routing do. When IP packets enter a network, they are delivered to the next AS as soon as possible.
Since most of the network neutrality debate is regarding Netflix, Comcast and Verizon these days, let’s see how hot potato routing works in this scenario. Netflix buys transit services from Cogent, another service provider (see our related white paper on peering relationships). Their agreement is that Cogent will deliver Netflix’s IP packets to their destinations. I, as a consumer, buy “Internet service” from Time Warner. My agreement with them is for all-I-can-eat Internet access with up to 30Mbps download speeds for about $80 a month. I expect to surf the web, watch videos on Netflix, iTunes, YouTube and other sites, and receive decent and reliable service.
When I watch a movie on Netflix, IP packets from Netflix are transited to Cogent and then from Cogent to Time Warner, then finally to the Apple TV in my living room. Time Warner and Cogent probably peer in many different locations. I live in California, and let’s say that the content I want to watch is stored in New York. Time Warner and Cogent peer in both locations. Whose network will transport the packets across the continent? Time Warner’s will, because hot potato routing will cause Cogent to pass the IP packets to Time Warner as early as possible. Timer Warner’s network will bear the heavier burden.
As we describe in our white paper, this is one of the main reasons why two top-tier service providers try to achieve symmetric traffic volumes in both directions before they decide to peer with each other settlement free. However, with video streams dominating the Internet traffic, it is becoming harder to achieve symmetry.
Would the situation change if we moved away from hot potato routing?
It depends on whether Time Warner, Comcast, Verizon and others want to get a piece of the over-the-top (OTT) revenue that Netflix is getting, or whether they are really trying to reduce their network congestion. Dropping hot potato routing would definitely reduce their burden.
My perspective – and many in Packet Design do not share this – is that I have paid Time Warner for the privilege of watching Netflix. If I am not paying them enough to cover their costs and profit margin, then they should increase their prices or put capacity limits on my connection. Yes, I would hate this as a consumer, but it is fair. It is a business relationship between us. I might consider switching to a different service provider, although in my case this is not viable, because at this time my only other choice is a 384Kbps DSL service.
The direct peering between Netflix and cable companies like Comcast and Time Warner, or the placing of Netflix caching servers in Comcast data centers, are business decisions between the two companies. Netflix is already paying for transit to Cogent; it can definitely split this payment to other SPs if it so wishes. However, forcing Netflix into direct peering and cutting out Cogent in the process sounds anti-competitive to me. We already have too little competition for consumers in some areas; should we also be cutting competition for enterprises?