Modeling Distance Decay Effects in Web Server Information Flows

摘要

Previous studies have shown that there is a direct correlation between the popularity of a medium and the cost in transmitting information through that medium. On this basis the World Wide Web may become more important as a mass medium than radio and television. Yet despite the enormous growth in Internet information and Internet-supported businesses, there has been, as yet, little in the way of quantitative spatial analysis of information flows on the WWW. One major issue is whether or not distance influences the web sites that a user may visit. If distance is a factor, then the location of servers delivering content becomes important. In this paper an analysis of information flows from U.K. academic web servers to the rest of the world is carried out. Using a UNIX utility called ping, the average time taken for a defined amount of information to travel between the United Kingdom and sixty-six other countries across the Internet was measured. This time measurement known as latency is used as a measure of distance on the Internet. The latency measurements are combined with counts of visitors from each of these countries to approximately one hundred U. K. academic WWW servers and used to build a simple gravity model of WWW information flows. The latency measurements between the United Kingdom and the other countries were gathered over a week in 1996. The counts of visitors relate to the total number of visits to the web servers over various time-scales for the years 1995 and 1996. We make the assumption that the distance measurements used are relevant to the visit counts. We also assume that all visitors from the .edu domain are geographically located in the United States. The gravity model is used to determine the effect of Internet distance on the number of expected visits to a web server. The study shows that latency values are a useful metric for measuring Internet distance. The results also demonstrate that the number of visitors to a web site falls off with distance on the Internet, as measured by latency values.