The Stanford lectures:
Over the past week as I engaged with the course I kept my eyes open for news that related to the fundamentals of the internet. There was no shortage. Stanford has a five-session continuing education course on iTunes U that I’d like to use as the kernel of this entry. The economics that were demonstrated went a long way toward helping me follow the logic of modern industry visionaries such as PBS’s Robert X. Cringely. It is impossible to sum up an 6 and a half hour course in a blog post. Still, I hope to give you the flavor and a bit of value-added reflection using momentous (maybe) current events.
The course is presented by the erudite internet researcher Ramesh Johari. He has published several articles on the counterintuitive network economics that underlay our information-based world. The lecture revolves around five uniquely positioned businesses–AT&T, Google, Akamai, NetFlix and Comcast. Each has a very different role and very different interests. Google and NetFlix are content looking to get bandwidth-heavy things like movies to the “eyes”; Comcast is “eyes”–they’re a major broadband provider; AT&T represents the backbone providers; and Akamai holds the omniscient middleman role of the Content Distribution Network (CDN). Each player has very different values and the question is always ‘who should pay who?’ and ‘what kinds of billing does the technology allow?’.
The phone telephony ‘network’ of copper was different from the protocols underlying the internet. As the linked video featuring one of the founders of the modern network describes, the components on the phone network are almost completely unreliable even with gold plated everything so it required that they have massive redundancy–if any part of the path went down, it was dead. If someone nuked the US, huge swaths of the phone network would be nonfunctional. Internet is different, internet routers know the nearest few connections and which direction will take you closer to your target for the lowest cost. If something breaks, take a different, somewhat more expensive, path.
In the first two sessions of the course, Ramesh covers the background necessary to understand the important protocols: stupid-simple network routing using IP; reliability brokered by TCP; and sheer speed established with UDP. He also hints to economic gotchas for those of us who are already enlightened to enjoy. For me, one of the most interesting concepts he introduced in the first lecture was the Internet Exchange Point. This location, often run by a nonprofit, allows a bazaar of internet traffic where everything is free, given that you’ve strung your own set of the wires into the joint to share. The wikipedia article lists hundreds of these. They are important. Without them developing countries would route all their traffic out to their ISP in the United States and back whenever you wanted to connect with the other ISP in town. Even Tanzania has one.
The course considers Network Neutrality consistently. As recent studies have found no differentiation between services means that to have services that require low-latency you need much more capacity than you might otherwise. Implementing this differentiation however necessarily enables and even requires profit/billing models that largely do not exist today. You can’t allow liers to monopolize a special bracket for high priority network traffic or it gets saturated. Ramesh points out a major niggle though: internet backbone creation funding model is really not stable yet. The last build-out was funded by the dot-com bubble and the hold-harmless bankruptcies that ensued. Over the next year it will largely be at capacity (or at least ‘lit up’).
Finally in the third episode he breaks it all open, He shows how the biggest internet companies (AT&T,Verizon) are actually stuck and will never have full control of the internet. A middleman company called Akamai that is “neutral” actually has more power over the internet than anyone else. Google can make all of their connection fees go away just by getting bigger and giving stuff away. This guy does research on the economic situation and he points out all the shortsighted moves of AT&T (internet backbone) and SBC (broadband) when they merged last year because they didn’t understand their own internet market and economics. They actually sold Cingular and then bought it back.
See Also:
In this Google Lecture Van Jacobson suggests that distributed protocols like BitTorrent and Avalanche are the latest abstraction which like the internet protocol upon phone infrastructure, will become the dominant network perspective for the coming years. He starts with an account of how the internet was invented from the groundwork of ma-bell and works his way to BitTorrent and the revolution in thinking.
Recent Comments