Linked

In 2003 Albert Laszlo Barabasi published a book on networking called 'Linked'. This book presented Barabasi's early insights on social, biological, business and computer networks. Barabasi had new sources of data to base his network research on - the structure of the Internet, and the linkages in the world wide web. 

Barabasi's research was a work of discovery. Of course there are rules defined by the IETF and the original developers of the Internet, but rules are made to be broken, especially when money is involved. How much had orthodox purity made way for secular pragmatism. Or in less philosophical terms, how had the collective demands of billions of end-users, the content created by millions and the incremental network expansions of thousands of service providers and network operators driven the evolution of the Internet?

The text book Internet is a mesh of routers. Hosts connect up to the nearest router, and data is forwarded hop-by-hop through other routers to its ultimate destination. The routers use links, in much the same way that we make use of roads and Interstates when we drive our cars cross-country.  Barabasi discovered the text book view of the Internet is wrong - Connectivity in the Internet is a lot more interesting than its architects had anticipated.  

To start, the map of connections between the routers in the Internet resembles more an Airline Map than the Interstate Highway map.  Highways provide connections to neighboring cities.  Routers connect anywhere in the world. If a lot of traffic is going from Chicago to LA, it makes sense to install a direct link between these two cities.  O-Hare and LAX have replaced Route-66.  How?  The direct links are created using optical networks.  

The Route-66 text book view of the Internet was always a myth.  The first nodes in the Internet were at UCLA, Stanford, Santa Barbara, Utah, and at BBN in Massachusetts.  Back then the links were created using old-fashion telecom technology. This layering of network technologies continues, but the technologies that are used in the different layers have all evolved.  Transport technologies are in a supporting role. They come and go based on relative economics. As the network grows to include more endpoints, new service delivery modes, and more, higher capacity links, the cast of characters in the supporting roles change.  Yesterday ATM and SONET, today Ethernet, OTN and DWDM. 

The Internet is bigger today than it was when Barabasi wrote Linked in 2003. The Atlas Internet Observatory 2009 Annual Report  shows that the Internet in 2009 is even more dependent on optical networking today than it was in 2003. Internet Transit revenues are down, replaced by disintermediation using optical technology. Fewer more highly connected ASNs source more of the data - just 150 ASNs account for more than 50% of Internet traffic. CDNs run their own optical networks to distribute content cheaper and quicker. Wholesale connectivity is a more significant business for wireline operators, replacing retail narrowband voice and data service revenues. Low latency, high bandwidth optical links provide the connectivity. The traffic on those links is predominantly IP. The client interface is predominantly Ethernet.

Networks are and always will be layered and multi-technology.  Layering provides choice, and control over performance and cost. 

Today's technology allows us to save by integrating the different transport technologies into common hardware platforms.  Software helps human operators understand the relationships between the different technologies, and empowers better decisions.