MobiTest : an evaluation infrastructure for mobile distributed applications

Abstract

Sophisticated applications that run on mobile devices have become commonplace. Within the wide realm of mobile software applications there exists a significant number that make use of networking in some form. Unfortunately, such distributed mobile applications are inherently difficult to evaluate. Conventional evaluations of such distributed applications are limited to small, real-world deployments consisting of, perhaps, a handful of phones. Such tests often do not have the requisite number of users to produce the desired performance. Also, these experiments do not scale and are not repeatable. To address all these issues, we sought to evaluate distributed applications in a virtual environment. Besides being cheaper, such evaluations are reproducible and scale significantly better. This thesis documents our efforts in working towards this goal. We discuss the designs that we iterated through, along with the problems we faced in each of them. We hope these problems will inform future designs that can solve the challenges that we weren't able to solve efficiently.