Let me offer a different, and perhaps, unpopular take on the current Comcast/Level 3 imbroglio. The conventional wisdom amongst Internet users is that Comcast is evil; therefore, Level 3 must be the innocent victim of Comcast’s capricious greed. I’m the first person to point out some real problems with Comcast; its secretive take on “network management” and ham-handed response to the regulators has made it no friends. However, this is a complex situation without clear-cut heroes or villains — in the network game, the Comcast/Level 3 conflict is business as usual.
First, let’s dismiss the alarmism . No one is “breaking the Internet.” That’s sort of the first rule among network engineers and peering managers, and a great deal of effort goes into meeting this goal. (Stacey did a good job explaining the issues involved in her post.) A peering dispute of this sort can lead to a partition — a condition where some Internet users can’t reach others — but that’s rare and considered to be highly reckless by the network operations community. While Cogent has been willing to prosecute its goals by partition, most other providers find it to be anathema.
So, what happened between Comcast and Level 3? Prior to this dustup, Level 3 was one of Comcast’s Internet transit providers. Level 3 sold wholesale Internet capacity to Comcast (as does Tata). Comcast engages in settlement-free peering with many networks, but as it’s not 100-percent peered, it needs upstream transit providers, such as Level3 or Tata , to offer a route for all non-peered traffic.
Level 3 has recently become one (but not the only, by any means) content delivery network for Netflix, which has radically increased the outbound traffic from Level 3 to Comcast, threatening to saturate the transit links between the two providers. That’s just business: no harm, no foul. But what Comcast found troubling was that Level 3 now looked a lot less like an upstream transit provider, and a lot more like a content delivery network (CDN). This is an issue, because Akamai and Limelight pay Comcast to deliver their traffic to their end users. Level 3, on the other hand, is paid by Comcast, ostensibly for transit, but now, seemingly, to deliver traffic that Level 3 has already been paid for — by Netflix. Level 3 wanted to get paid both ways: coming and going. That’s nice work if you can get it.
Level 3’s transformation into a CDN raised additional questions. The era of the Internet backbone providers has been coming to a close. The Level 3s, Global Crossings, and such are in some ways relics; the real power is with the last-mile carriers — cable, DSL, FTTP providers — and the content providers — Google, Microsoft, Akamai, Limelight, et al. These eyeball and content providers have been increasingly disintermediating the classical Internet backbone providers out of the equation by interconnecting amongst themselves. While providers like Level 3 still have some utility there, that utility is shrinking rapidly.
Comcast and its EVP of Networks, John Schanz, have surely been asking themselves, “Why have we been paying Level 3 for transit?” After all, Comcast is a much larger and more important network, especially in terms of traffic, in the general scheme of things. While many of the old backbones have been acquired — and more will be in the coming year — Level 3’s titanic debt load has scared off just about everyone, leaving it high and dry.
All of this said, no one here did anything wrong; these disputes aren’t uncommon. Well, no one did anything wrong until Level 3 whined to the press and regulators. When you’re on the losing side of a peering dispute, you’re always tempted to complain to regulators about how you’ve been treated unfairly. Then you remember a regulated peering arrangement is the only thing worse than what we have now. More regulation puts more power in the hands of the big providers, not less, as they have the experience, resources, and the legions of regulatory lawyers necessary to “play the game.”
Look at the Telecom Act of 1996; the Baby Bells won that fight due to their superior regulatory game-playing ability. The cable companies, PTTs (national telephone companies), AT&T, and Verizon would surely come out on top of any major movement towards peering regulation which would cut out the large number of small firms that peer 50 percent to 70 percent of their traffic.
So, how does this play with network neutrality? The short answer is that it’s completely unrelated. Netflix isn’t paying a dime to Comcast. Level 3, which sells services to Comcast, is. It’s the price of doing business when you have a large amount of outbound traffic. No user will pay more to watch Netflix movies. Netflix certainly won’t see this passed on to them; the low cost of Level 3’s service is what won the deal, and if that goes away, so will Netflix, to greener pastures. Instead, this is a battle where a provider that does no business with consumers (Level 3) will pay money to a company that does provide services to consumers (Comcast). There will be no prioritization of traffic as part of this deal, only delivery.
This in no way absolves Comcast of its duty — as good corporate citizens, not to mention a company trying to make a merger happen — to be more transparent, to deal openly with its customers, to improve its levels of service, to provide competitive prices. These are all areas where Comcast has a lot of work to do, to put it mildly. However, in this latest peering conflict, one of the dozens that I’ve observed or participated in, there is nothing malevolent or damaging to consumers, on either party’s behalf.
Daniel Golding is Managing Director of DH Capital, a boutique investment firm in the Internet infrastructure and datacenter sector. Prior to that, he worked at Tier1 Research, was the Global Peering Manager at AOL. Daniel is an engineer with 20 years of network and mission critical datacenter experience. Disclosure: He takes no money from CDNs, Netflix, Comcast, Level3, or anyone else involved in this story, but in the past, he has sold research to those companies.
Related GigaOM Pro Content (sub req’d):