Abstract
Modern mobile devices, such as smartphones and tablets, are offering increasingly complex functionalities. We use them daily in activities ranging from entertainment, to solving professional tasks. Although mobile devices are growing in functionality and computing power, we believe the role of more powerful infrastructure, to augment the capabilities of mobiles, will increase. Surveying the state of the art, we can see that developers and researchers alike study ways of accessing, from the mobile device, resources from the Cloud, from community resource pools, and from personal networks.
In this work, we focus on defining an exploratory space for offloading concerns in the context of applications for mobile devices, on identifying novel offloading mechanisms, and on conducting experimental evaluation for both. First, we survey existing research efforts regarding offloading for mobile devices and propose a General Offloading Model and a Taxonomy for Offloading Concerns. We also propose an Exploratory Space that shows novel offloading mechanisms, as well as the opportunity of a design space exploration. Second, we present our efforts to investigate Communication Adaptation and Offloading for applications spanning the mobile device and custom hardware extensions, and to investigate Computation Adaptation and Offloading for loop-based applications. Third, we summarize our efforts to collect traces for thousands of Facebook games and hundreds of native mobile applications. We also propose a workload model, for both macro-level information---the number of users---and micro-level information---the operations users trigger, that we use to characterize and model workloads of a large number of applications and to generate synthetic traces. Fourth, we design and implement an Integrated Offloading System to conduct design space exploration and to experiment with various novel offloading mechanisms. Fifth, we show the experimental results for the empirical evaluation of the Integrated Offloading System, using both real and synthetic traces.
The main contributions of our work are the workload model and the offloading system. Our workload model for online social applications triggered positive feedback from the research community, leading us to plan further improvements, that will refine the model and allow it to cover more applications, with greater fidelity. The integrated offloading system can continue to serve as basis for further empirical evaluation, both with real applications and simulations.
Download Thesis Download Summary (RO)
Timeline
Sep.2013
Jul.2013-Aug.2013
Oct.2012-Mar.2013
Apr.2012
Nov.2011-Feb.2012
Oct.2011
Oct.2010
Internal Defense at the "A&C Ph.D. Sudent Days 2013"
Internship at TUDelft: Integrated Offloading System
Internship at TUDelft: Towards a Workload Model for Online Social Applications
Started collaboration with Alex Iosup (TUDelft) on Social Gaming and Mobile Offloading
Internship at HTW-Dresden: Enabling Mobile Devices for Home Automation using ZigBee
Defined the topic: Extending Capabilities of Mobile Devices through Cloud Offloading
Started PhD in Mobile and Cloud Computing