|Keywords:||Cellular Data Usage Reduction; Data Offloading; Mobile Computing; WiFi Data Usage Improvement; Electrical engineering; Computer engineering|
|Full text PDF:||http://arks.princeton.edu/ark:/88435/dsp014f16c505p|
Worldwide, mobile data usage has been increasing dramatically. As a result, users data usage costs increase. Mobile data offloading to WiFi where available could greatly decrease the usage of cellular data networks and reduce users data usage costs. Challenges arise, however, in planning how and when to exploit WiFi versus cellular connectivity. In this thesis, I first develop an optimal MILP-based scheduling framework to explore the benefits of delay-tolerant WiFi offloading. Assuming perfect knowledge of future network and data usage characteristics, the proposed framework finds minimum-cost (in terms of cellular network data) schedules for multiple application data streams with varying size and delay tolerance, communicating via networks with varying coverage and bandwidth. Even though a MILP-based approach has great potential, the proposed technique relies heavily on the predictions of network conditions and data usage. This thesis also proposes efficient, easy-to-implement heuristic approaches which rely on less prediction. Furthermore, my thesis prototypes the offloading framework on an Android smartphone including scheduling, delay tolerance, and seamless switching features. Overall, delay-tolerant techniques average more than 2X reduction in cellular data usage, and for some scenarios, the reduction is as high as 5X. While exploiting delay tolerance offers significant energy and cost benefits, a key question remains: how long to wait? Prior work does not discuss how to estimate application delay tolerance without explicit help from programmers, nor how to adjust the estimate dynamically. I propose and implement statistical and heuristic decision techniques which use small hints (i.e. metadata) to catch users data access request and use these request patterns to deduce an applications delay tolerance dynamically. Experiments show that dynamically adaptive decision schemes achieve up to 15% further cellular data reduction compared to fixed static delay tolerance values. To summarize, this thesis proposes and evaluates both optimal and close-to optimal techniques and provides the required mechanism support for minimizing cellular data usage while exploiting applications delay tolerance dynamically. Overall, my thesis offers insights for real-world implementations of such offloading solutions by prototyping practical connectivity optimizers, and considering a range of design alternatives.