CrowdMeter: an emulation platform for performance evaluation of crowd-sensing applications.

ABSTRACTIn this paper we introduce CrowdMeter, an emulation platform for predicting performance of large-scale crowd-sensing applications. CrowdMeter architecture follows natural decomposition of the application under evaluation and provides features for emulation of mobile devices and access network links. It leverages virtualization and cloud-based infrastructure-as-service resources to offer necessary scaling. We instantiate CrowdMeter architecture using off-the-shelf components and public-cloud resources, and perform a preliminary evaluation of its emulation fidelity focused on the communication services. The results confirm that CrowdMeter can successfully capture important aspects of real-world performance of different wireless access links. They also illustrate the ease-of-use and the scalability of the platform in terms of number of emulated mobile devices.