Section: Application Domains

Quality of Experience

Understanding how users react to different levels of network performance presents two main challenges:

  1. User perception is subjective and contextual. Different users may have different tolerance levels to network performance and the same user may have different expectations under different circumstances. Take for example the round-trip time (RTT), a typical network performance metric. If RTTs are larger 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 has a playout buffer to mask some network delay). Take another example of a user downloading her email. This user may tolerate some delay when she is leisurely checking her email at home, but she may become extremely frustrated with the same delay if she is in an airplane and needs to download her email just before takeoff.

  2. It is challenging to “measure” users. We must develop methods to measure the user perception of network performance as users perform their routine online tasks. It is hence important that these methods are not too intrusive. Otherwise, users are unlikely to participate in the experiment. In addition, we must capture user perception at different levels of performance and in a variety of scenarios.

We will develop tools that run on end systems to collect network performance data annotated with the user perception. These tools will adopt a hybrid measurement methodology that combines network measurement techniques to infer application performance with techniques from HCI to measure user perception. We will later use the resulting datasets to build models of user perception of network performance based only on data that we can obtain automatically from the user device or from user's traffic observed in the network. Models of user perception of network performance can be used to detect when performance is poor to trigger diagnosis or to adapt network/application performance to better serve users.