Understanding end-user perception of network problems

Abstract

It is widely assumed that certain network characteristics cause end-user irritation with network performance. These assumptions then drive the selection of quality of service parameters or the goals of adaptive systems. We have developed a methodology and toolchain, SoylentLogger, that employs user studies to empirically investigate such assumptions. SoylentLogger collects client-centric network measurement data that is labeled by the end-user as being associated with irritation at perceived network performance (or not). The data collection and labeling occurs in real-time as the user normally uses the network. We conducted a study that tracked 32 ordinary users over a period of 3 weeks and then used that data to test common assumptions about network sources of user irritation.