Paper 3
P-Bench: Benchmarking in Data-Centric Pervasive Application DevelopmentAuthors: Sabina Surdu, Yann Gripay, Vasile-Marian Scuturici, and Jean-Marc Petit |
AbstractDeveloping complex data-centric applications, which man- age intricate interactions between distributed and heterogeneous entities from pervasive environments, is a tedious task. In this paper we pursue the difficult objective of assessing the ”easiness” of data-centric devel- opment in pervasive environments, which turns out to be much more challenging than simply measuring execution times in performance anal- yses and requires highly qualified programmers.We introduce P-Bench, a benchmark that comparatively evaluates the easiness of development us- ing three types of systems: (1) the Microsoft StreamInsight unmodified Data Stream Management System, LINQ and C#, (2) the StreamIn- sight++ ad hoc framework, an enriched version of StreamInsight, that meets pervasive application requirements, and (3) our SoCQ system, de- signed for managing data, streams and services in a unified manner. We define five tasks that we implement in the analyzed systems, based on core needs for pervasive application development. To evaluate the tasks’ implementations, we introduce a set of metrics and provide the exper- imental results. Our study allows differentiating between the proposed types of systems based on their strengths and weaknesses when building pervasive applications. |