Beyond Broadband Access: Developing Data-Based Information Policy Strategies

CHAPTER 3 Broadband Microfoundations

51 CHAPTER 3 Broadband Microfoundations The Need for Traffic Data steven bauer, david clark, and william lehr To date, most of the empirical effort to understand broadband service markets has focused on availability and adoption metrics and data. Data of this sort is indeed valuable when the dominant policy questions concern penetration and uptake. However, as broadband availability and penetration saturate, such data will become less informative. The next set of questions, both for service providers and regulators, will center on the continued health of the broadband access market: levels of investment, the competitive landscape , the evolving definition of broadband, the degree of neutrality in consumer access, and the nature of interconnection among providers. Our position is that network traffic data will be central to understanding and answering many of these questions. To answer these sorts of questions it will be helpful to know such things as the distribution of usage across the user population, the characteristics of users that participate during peak periods of network congestion, and the variance in usage and how it differs by type of user. This data will help inform forecasts of capacity/infrastructure investment needs (e.g., how much bandwidth does a subscriber need? How much sharing is feasible at which points in the network?), to understand ISP costs, and to assess network management practices (e.g., traffic engineering). Better traffic data will provide insights into consumer adoption decisions and the evaluation of product offerings (e.g., how important are peak rates versus average data rates?). s. bauer, d. cl ark, and w. lehr 52 background The Internet is often compared to the network of highways, streets, and roads that make up the transportation system. Both are vital infrastructures that provide businesses with access to materials and markets, and provide people with access to goods, services, recreation, jobs, and each other. For transportation networks, it is generally recognized that traffic data (i.e., the volume of traffic, congestion information, incident reports, and so on) is as important to understanding the state of the network as is information about where the roads or links actually are. The same is true for the Internet. In transportation networks, traffic data is valuable over both short time scales (e.g., allowing real-time traffic management to reroute commuters around a rush hour accident) and over longer time scales (e.g., for planning maintenance cycles and capacity expansion investments). During periods of congestion,1 traffic data and real-time traffic management via lights, tolls, and special commuter lanes has proved especially important in enabling more efficient utilization of the existing transportation infrastructure. Improving the efficiency of the existing infrastructure delivers benefits in the form of reduced commute times (contributing directly to labor productivity ), improved safety, and reduced pollutant emissions through intelligent traffic management policies.2 On the Internet, traffic data is similarly important to network operations. Over short time scales ranging from less than a second to hours or days, traf- fic data is an input into systems (both automated and human-centered) that make routing decisions (e.g., balancing loads across different network links), identify suspected or actual security or transmission failures, and implement traffic management policies.3 Over longer time scales measuring months or years, traffic data is vital to capacity planning and provisioning, allowing capacity to be efficiently installed in advance of demand, thereby better accommodating future traffic growth without congestion-related disruptions . Thus traffic data is essential to almost all the practical dimensions of network management and to the political, regulatory, and theoretical questions of what constitutes good, acceptable, or socially desirable network management. While traffic conditions on the highway and roadways can be observed externally (via both technical sensors and human observations), information about Internet traffic and the level of congestion of the different autonomous networks that collectively compose the Internet is limited. While individual network operators generally have a good idea about the state of their own networks, outside stakeholders have little visibility into the state of traffic on networks. Networks can be probed and tested by outside observers to derive some measurements, but the scope and confidence of such measurements is broad ban d mi cro fo un d ati o n s 53 limited compared to the accuracy and breadth of information available to...