What makes modern Internetworking hard to cover is that you have to actually understand a fair bit about the underlying technology and economics to make sense of it. This turns out to be difficult for the press, public and politicians. A recent Light Reading article includes an amusing quote attributed to Chris Sacca, Google’s head of special initiatives:
“We have one peering point in San Francisco and some journalists say that we’re trying to take over the world,” Sacca says. “That’s the thing that a lot of journalists don’t get,” he says, “is that one peering point does not a telecommunications network make.”
This was said in the context of a story about Google’s use (or not) of dark fiber and the relationship between Google and AT&T. Now, either the GOOG has a much smaller network infrastructure than just about everyone thinks it does, or this quote isn’t saying what it appears to be saying. I’m not blaming Mark Sullivan, the intrepid Lightreading reporter who wrote the story, for getting it wrong. I do think it’s inconceivable that Sacca said and meant this, so there must have been some serious miscommunication. Topics about wide-area networking and peering are sophisticated and hard to talk about in plain language, both for the interviewee and for the reporter. But the confusion and misunderstandings are doing nothing to help the current national debate about net-neutrality, which is obviously an important topic that many people in the US care about.
You may be asking yourself: what does this story about Google and AT&T have to do with net-neutrality? I’ll take a stab at an answer.
Let’s be clear:
- Google has more than one peering location.
- Google does not peer with AT&T at all.
- Google doesn’t even peer in San Francisco, technically (San Jose, yes, but not SF).
Google has dozens of network points of presence around the world where they exchange traffic. I don’t actually see evidence that Google (AS15169) interconnects with AT&T (AS7018) at all, when looking in Renesys’s Routing Intelligence application and the underlying routing data. So when the article says “The part of the network neutrality debate that is never heard, Sacca says, is the fact that Google and AT&T have a massive peering arrangement,”, it is hard to know what that means. It’s possible that Sacca is referring to the SBCIS network (AS7132), rather than the AT&T classic network. In any case, if Google and AT&T do interconnect in a way that is not visible in Renesys’s routing data (which is possible), there is zero chance that it is at a single location. Grown up networks just don’t interconnect like that.
Google runs a really big network. It’s not the same kind of network that a carrier like AT&T runs, but it is a network, nonetheless. Google and AT&T both design and build networks to meet the business needs of their corporations. In the case of AT&T, that means that the network must reach the enterprise and service provider end-users that need connectivity, it must aggregate traffic into backbones effectively and must take that traffic to other networks, including customer networks and peer networks. In the case of Google, it means that the network must interconnect them to the rest of the Internet in order to efficiently deliver the web pages consumed by the constant crawl of the web, and to efficiently deliver Google’s content (search results, gmail pages, ads, news, videos) to Google’s user base. All of this is really pretty straightforward.
A network of Google’s size cannot interconnect with anyone, much less a network of AT&T’s size at a single point. Google generates and sinks traffic to a number of different locations. It does this for reliability reasons (multiple data centers), for performance reasons (servers closer to the user respond faster), for business reasons (people who operate the networks that connect Google’s customers require multiple points of interconnection). Google also needs to handle periodic outages and maintenance to its network and server infrastructure in order to maintain 24×7 service availability in the face of near-constant failure of individual components (when you run a lot of anything: servers, routers, circuits, something is always broken).
So back to net neutrality and the politics behind all of this: ultimately, every conversation about AT&T and Google right now is also about net-neutrality. AT&T and Google are the poster children for the two sides of the net neutrality debate. AT&T wants the right to reduce the quality of some traffic on its networks travelling to its subscribers so that it can offer premium versions of the same services and make more money. Google depends upon AT&T (and Verizon and other access networks) to deliver its revenue-producing ads to end-users and doesn’t want the traffic restricted or hampered in any way. This has all been covered pretty extensively in the mainstream press, the computer press, and the business press.
But even with extensive coverage and near-constant debate on the subject of net-neutrality, the discussion has ultimately provided very little clarity. Frequently the participants in the debate are simply talking about different things, or referring so imprecisely as to leave entirely in doubt what the subject of their position may be. And that’s the problem. Without a common base of facts and understanding about the technology, law, history and business models, it is impossible to have a useful discussion about network neutrality. Which is a problem, because it turns out to be a conversation that many people are desperately trying (and failing) to have.
Ultimately, that’s what caught my attention about this article. I’m not sure I care that much about how much dark fiber Google has rights to and exactly what they plan to do with it (why people care so much about this and other Google-related topics is the subject for a future post coming soon to this space). But I do think that we ought to be able to have a public discourse about these topics from a roughly common fact base and frame of reference. I hope we’re getting closer.