A Status Report on Internet Measurement

It is an oft-stated goal that policy-making be evidence-based. In the context of the Internet, evidence can take many forms. One important source is measurements of the technical character of the network. A vigorous community of academic and other third-party researchers carry out measurements on the Internet, but this community is not well-connected to the Internet policy community. In this paper we undertake a bridge-building effort between these two communities, drawing on a review of several recent technical conferences and workshops on Internet measurement. We summarize three related areas: research questions the measurement community is pursuing, the range of research methods used and key research results. While network neutrality seems still to be a major focus of the policy community, it never became a central focus of Internet measurement research, for several reasons. Most importantly, it is not an area amenable to rigorous scientific inquiry, since terminology is itself not rigorous, e.g., "reasonable network management." More practically, current discrimination is more likely to have an economic rather than a technical form, e.g., zero rating or discriminatory pricing of interconnection; evaluating these types of discrimination requires different sorts of data and analysis. The technical measurement community has a broader range of interests. This paper will review several measurement projects published this year, including: the extent to which end-to-end encryption is being subverted by manipulation of the Certificate Authority system (for both malicious and arguably benign purposes); utilization of the IPv4 address space and migration to IPv6; the nature of network outages; the utilization and implications of Facebook's Free Basics service, the prevalence of ad-blockers; the prevalence of Carrier-grade NATs; how applications evade firewall blocking; and DNS and BGP behavior. We will also report on measurement efforts looking at specific aspects of "reasonable network management": treatment of traffic in mobile networks and measurement of interconnection links between ISPs. These studies targeted a technical audience. Although sometimes they identify policy implications of their results, it is mostly left as an exercise for the reader. We will attempt to bring out some of those issues. Another interesting aspect of our meta-analysis is the wide range of measurement methods, including: active probing (sending specific packets and observing results); instrumentation of applications (both at client and server); exploiting data initially gathered for other purposes; connecting to network control mechanisms such as routing protocols; and the creative use of ads placed on web sites to trigger remote probing. Our goal with this paper is to inform the policy community of the diverse work being done by the technical measurement community, and to stimulate interest in new sorts of collaboration between technology and policy researchers