Date of Award
Doctor of Philosophy in Computing Sciences - (Ph.D.)
Adriana Ioana Iamnitchi
Since people generate and access most digital content from mobile devices, novel innovative mobile apps and services are possible. Most people are interested in sharing this content with communities defined by friendship, similar interests, or geography in exchange for valuable services from these innovative apps. At the same time, they want to own and control their content. Collaborative mobile computing is an ideal choice for this situation. However, due to the distributed nature of this computing environment and the limited resources on mobile devices, maintaining content availability and storage fairness as well as providing efficient programming frameworks are challenging.
This dissertation explores several techniques to improve these shortcomings of collaborative mobile computing platforms. First, it proposes a medley of three techniques into one system, MobiStore, that offers content availability in mobile peer-to-peer networks: topology maintenance with robust connectivity, structural reorientation based on the current state of the network, and gossip-based hierarchical updates. Experimental results showed that MobiStore outperforms a state-of-the-art comparison system in terms of content availability and resource usage fairness.
Next, the dissertation explores the usage of social relationship properties (i.e., network centrality) to improve the fairness of resource allocation for collaborative computing in peer-to-peer online social networks. The challenge is how to provide fairness in content replication for P2P-OSN, given that the peers in these networks exchange information only with one-hop neighbors. The proposed solution provides fairness by selecting the peers to replicate content based on their potential to introduce the storage skewness, which is determined from their structural properties in the network. The proposed solution, Philia, achieves higher content availability and storage fairness than several comparison systems.
The dissertation concludes with a high-level distributed programming model, which efficiently uses computing resources on a cloud-assisted, collaborative mobile computing platform. This platform pairs mobile devices with virtual machines (VMs) in the cloud for increased execution performance and availability. On such a platform, two important challenges arise: first, pairing the two computing entities into a seamless computation, communication, and storage unit; and second, using the computing resources in a cost-effective way. This dissertation proposes Moitree, a distributed programming model and middleware that translates high-level programming constructs into events and provides the illusion of a single computing entity over the mobile-VM pairs. From programmers’ viewpoint, the Moitree API models user collaborations into dynamic groups formed over location, time, or social hierarchies. Experimental results from a prototype implementation show that Moitree is scalable, suitable for real-time apps, and can improve the performance of collaborating apps regarding latency and energy consumption.
Khan, Mohammad A., "Context-aware collaborative storage and programming for mobile users" (2016). Dissertations. 95.