Scientific objectives

In the BETTERNET project, we will deliver methods, algorithms, and software systems to evaluate the quality of Internet end-to-end communications in terms that are more meaningful to users. More specifically, the scientific and technical objectives of BETTERNET are:

  • A more comprehensive definition of quality of Internet access that combines a set of metrics that go beyond traditional network performance metrics to better capture user QoE by considering end-to-end communications;
  • Algorithms to measure the quality of Internet communication and infer traffic shaping policies and practices (e.g., diverting traffic to proxies and specific servers) of stakeholders. Given our methods are to be used for regulatory purposes, it is key that ISPs, users, and regulators trust the resulting inferences;
  • Algorithms to infer whether the ISP, the CSP, or the user local environment are involved in the identified performance issues. As no single actor is in a position to clearly state and diagnose, deterministically or probabilistically, the origin of the degradation, traditional techniques will not reveal the root cause. It is essential to discern, separate sources of degradations through distributed and diversified passive and active measurement techniques, as well as advanced measurement and probing strategies;
  • Techniques to profile Internet usage patterns. BETTERNET will analyze Internet usage patterns both from network performance and sociological point of view to gain new insights on user online activities and configurations (devices, networks. . . );
  • Methods to assist users to select the best fitting Internet service plan and configuration. The BETTERNET project will develop algorithms to predict not only which service users should buy given their usage patterns but also how to configure these services to get better performance. We aim at offering solutions to estimate/pict service delivery quality relying on measurements at the network level. This will be a key asset to leverage the efforts of regulation instances, as ARCEP (Autorité de Régulation des Communications Électroniques et des Postes), that try to find the best representation of Internet access offers without analyzing the whole chain of intermediate elements;
  • A system that runs on end-devices that can continuously and adaptively probe the quality of Internet end-to-end communications.

To achieve these objectives BETTERNET must overcome three main challenges:

  1. The mapping of Quality of Service (QoS) to QoE requires models of user perception of network performance, which is often subjective and contextual. Different users may have different tolerance levels to network performance and the same user may have different expectations under different circumstances. For example, if delays are higher than usual, a user who is doing remote login may feel that the connection is unusable, whereas another who is watching YouTube may notice no problem (because YouTube uses a playout buffer to mask some network delay).
  1. Users local network environment and usage profiles are heterogeneous. The set of devices, configurations, and access technology in users’ local network as well as the set of applications users use vary significantly from one user to another. We must develop sophisticated techniques that can learn and adapt to any local environment and usage profile.
  2. Internet services are complex. Performance issues can appear at different layers of the protocol stack and at different components of the network or Internet services. Our algorithms must adapt to the large variety of Internet services and network configurations.

Comments are closed.