The absence of a standard measurement plane, which provides performance and comparison information, raises a large number of challenges for a general understanding and enhancement of the Internet Service Accesss:

Assessment of the quality of the Internet access

There has been considerable interest in methods and tools for measuring the performance of both residential and mobile Internet access. However, current techniques mainly focus on measuring the performance of the last-mile link, neglecting other elements, such as the home network, proxies and connectivity to Content Delivery Network (CDN), which are also critical to determine a user’s QoE. Second, current techniques mostly rely on probes connecting to one single target server, which can introduce bias related to the placement and performance of this server. Hence, one must diversify the set of measurement targets in space and time and across protocols and service providers. Third, current techniques mainly use objective network performance metrics (for example, capacity and delay of the access link), but this notion of performance, although relevant and important, is unsatisfying to users. Users rather care that their web pages load fast and that their YouTube videos do not stall. Thus, new metrics as the Page Load Time (as defined by W3C1) have to be evaluated.

Diagnosing end-to-end communications

When users experience performance or connectivity problems, there is little they can do. The most common attempt to solve the problem is to reboot or call the provider’s hot line. Network providers have more information to use in diagnosing problems in their networks, but the complexity of today’s services makes it hard for ISPs to know what is degrading their customers’ experience. CSPs also monitor the quality of end-to-end paths to deliver the best performance to users. Despite the recent advances in end-host monitoring and troubleshooting, network performance and fault diagnosis are mostly manual and completely ad hoc. In this context, we must also detect whether any poor QoE is the result of one stakeholder deliberately throttling traffic and blocking some applications or services. There is currently no general technique that can detect all different types of traffic differentiation that can occur in practice.

Collecting user experience in the field

Collecting information from the perspective of end-users requires the development of solutions and algorithms to run on users’ devices without compromising the user experience (e.g., privacy, battery consumption). Current initiatives to map the quality of signals (OpenSignal2, Sensorly3, etc.) offer dedicated mobile applications for users to probe the performance of their Internet access and report on the metrics from their mobile devices. However, most of these approaches do not provide any incentive in order to raise a wider adoption of these applications and provide a critical coverage of the population and the territory. Moreover, these approaches are technologically limited to the delivery of a mobile application with little latitude on the types of metrics that can be contributed by the end-users.


Internet has evolved and cannot be reduced to communication and software elements. Users are now directly involved within this worldwide system. Thus, it is important to understand how people use the Internet and to track the emergence and differentiation of new collective practices. Services and data have a huge impact on society, but society and uses in turn highly influences the structures of the Internet. Even if the Internet measurement field has expanded, few public measurement studies have associated social and computer scientists when analyzing end-user practice with flow measurement or with anomalous machine behavior, or guess the business conflicts between main actors, or the evolution of the geography of the Internet over time.
Marketers, industrials and providers produce this kind of data, but they do not share their datasets and results with civil society and/or scientists. Today, ISPs mainly advertise the cost and some generic upload and download speeds, but speed alone cannot accurately capture a user’s online experience. Few studies have also faced the complexity of measuring the end-user connected devices configurations and its discriminating uses between social space, working time, transport, home or leisure, for example.
There is a recent interest in combining more traditional network and system performance measurements with Human-Computer Interaction (HCI) techniques that explicitly ask user feedback on network performance and faults. None of these early studies, however, has addressed the problem of profiling a user’s usage pattern and identifying suitable service plans. Combining measured data with quantitative and qualitative declarative data (obtained from traditional surveys and/or interviews) will help the social scientists to discover new fields of research in relation with the Internet (perception and cognition of Internet performance; sociology of cultural practices; sociology of Internet actors, etc.) and to structure collective data- sets for multiple sociological analytic purposes (behavioural analysis, routines and live styles, audience and economic models, appropriation and misappropriation, cultural habits. . . ).

Comments are closed.