Abstract
Capturing the flow of information between cities is a challenging task. Historically, flow analyses have focused on goods, capital, and people, all of which can serve as proxies for estimating the volume of spatial interaction between places. However, with the advent of the Internet and its ability to both facilitate and accelerate the exchange of information, it is somewhat surprising that so few studies have examined the geographical characteristics of Internet flows. Aside from the initial challenges associated in acquiring network flow data, there are additional constraints inhibiting such efforts, including privacy concerns, the geographic rectification of flows, and the ability to manage and visualize massive datasets. The purpose of this paper is to outline a basic methodology for capturing Internet flow data and to provide a brief empirical analysis of these data for the Internet2 network in the United States. Results suggest that asymmetries exist between ingress and egress connectivity and flows throughout the United States.
Notes
An Internet Protocol address is a unique, 32-bit number that serves as an address for a specific TCP/IP host on the Internet. For example, an IP address for Indiana University is 129.79.174.243.
Derudder and Witlox Citation(2005) note that many of the sources used for air transport analysis over-emphasize international flows (e.g., New York to London) versus “global” flows (e.g., New York to Los Angeles). There are also problems associated with determining the differences between flows generated by world city processes and those generated by non-affiliated factors, such as tourism.
For a thorough review, see Alderson et al., Citation(2006).
Geographical indeterminacy refers to issues associated with pinpointing Internet content production and consumption when dealing with large network providers. For example, while Comcast serves millions of customers throughout the United States with cable broadband, all IP addresses are registered to an address in Mt. Laurel, NJ.
Hints to location of servers are often given in the actual domain names reported by a traceroute query. For example, xe-1-0-0.0.rtr.atla.net.internet2.edu corresponds to an Internet2 router located in Atlanta, GA.
For a list of university members, see http://members.internet2.edu/university/universities.cfm
For a list of corporate members, see http://members.internet2.edu/corporate/corporate.cfm
For a list of affiliate members, see http://members.internet2.edu/affiliate/affiliates.cfm
UDP is part of the TCP/IP protocol suite that was created to provide a way for applications to access the connectionless features of IP. It is important to note that UDP allows for the exchange of packets without acknowledging or guaranteeing delivery (Newton, Citation2002).
For example, packets that are part of a web surfing session usually have a destination port of 80 and e-mail usually has a destination port of 25.
This assumes a constant utilization density.