Utility-based resource-aware framework for information caching and sharing in mobile and distributed systems

The Internet is evolving into an indispensable service delivery infrastructure and infinite information database. Along with the technology advancements in mobile and wireless networks, ubiquitous information service is becoming a reality in which users can access information anytime anywhere. However, the user mobility, network heterogeneity and resource constraints impose significant challenges to provide ubiquitous information services. In this dissertation, a utility based resource aware framework is proposed to enhance ubiquitous information availability to mobile users through data caching and peer-to-peer sharing. The framework considers the constrained resources of mobile and distributed environments and provides flexible, efficient and scalable data access services to the mobile users. The major contributions of this framework are as follows. First, we introduce a novel energy and bandwidth efficient data caching mechanism, called GreedyDual Least Utility (GD-LU), to enhance dynamic data availability to mobile users in cellular networks. Based on the utility function derived from our analytical model, we propose algorithms for cache replacement and passive prefetching of data objects. Second, we introduce a novel scheme called energy efficient peer-to-peer caching with optimal radius (EPCOR), to enable peer-to-peer information sharing in multi-hop hybrid networks. In EPCOR, a peer-to-peer overlay network is built among the mobile users to facilitate cooperative sharing of data based on network proximity and data preference. In order to conserve energy, each mobile user shares a data item in a cooperation zone. An algorithm is developed to determine the optimal radius of the cooperation zone. Third, we investigate location-aided information retrieval in large-scale mobile peer-to-peer (MP2P) networks. A novel scheme, called Proximity Regions for Caching in Cooperative MP2P Networks (PReCinCt) is designed to utilize location information to support scalable data retrieval. In the PReCinCt scheme, the network topology is divided into geographical regions where each region is responsible for a set of keys representing the data. Each key is then mapped to a location based on a geographical hash function. We evaluate and validate the developed algorithms both analytically and experimentally. We have conducted extensive experiments using large scale simulations to evaluate the performance of proposed framework. Our analytical and experimental results show that the framework can efficiently provide ubiquitous information services in mobile and distributed environments.