| Motivated by the intermittent connectivity that many mobile users experience, we have been investigating mechanisms to improve their access to data. We propose 7DS, a system that addresses the challenge of increasing data availability by providing a novel mechanism that enables wireless devices to share resources in a self-organizing manner, without the need for an infrastructure.; 7DS is an architecture, a set of protocols, and an implementation enabling resource sharing among peers that are not necessarily connected to the Internet. Peers can be either mobile or stationary. The focus is on three facets of cooperation, namely information sharing, bandwidth sharing, and message relaying. In the information sharing facet, peers query, discover, and disseminate information. Hosts acquire the information from the cache of other peers. For message relaying, hosts forward messages to the Internet (when they gain Internet access) on behalf of other hosts. We investigate the bandwidth sharing in wireless LANs and in video-on-demand servers. When bandwidth sharing is enabled in a wireless LAN, the system allows a host to temporarily act as an application-based gateway and share its connection to the Internet. In the video-on-demand case, the server shares dynamically its disk bandwidth among the clients.; The system adapts its communication behavior based on the availability of energy and bandwidth. For the information sharing and message relaying, we model several schemes depending on their type of cooperation among hosts, querying mechanism, energy conservation, host density, and transmission power. We evaluate these schemes and their impact on information discovery and data availability via simulations. We also provide an analytical model for a baseline scheme and show that the analytical results on data dissemination are consistent with the simulation results. For the case of bandwidth sharing in wireless LANs, we design a lightweight protocol and present its benefits via simulations. For the bandwidth sharing in a video-on-demand multi-disk server, we present novel retrieval techniques that take advantage of layered multimedia information and replication to dynamically reallocate the disk bandwidth. We model a multi-disk environment and show its performance in the case of no replication, partial replication and full replication as a function of user access skew. Our scheduling algorithm for the retrieval of streams can double the disk bandwidth utilization of the server. |
