Location

Wiesner Room

Description

In this thesis we lay the foundations for a distributed, community-based computing environment to tap the resources of a community to better perform some tasks, either computationally hard or economically prohibitive, or physically inconvenient, that one individual is unable to accomplish efficiently. We introduce community coding, where information systems meet social networks, to tackle some of the challenges in this new paradigm of community computation.
We design algorithms, protocols, and build system prototypes to demonstrate the power of community computation to better deal with reliability, scalability and security issues, which are the main challenges in many emerging community-computing environments, in several application scenarios such as community storage, cooperative offline Wikipedia and image sensing. For example, we develop a community storage system that is based upon a distributed P2P (peer-to-peer) storage paradigm, where we take an array of small, periodically accessible, individual computers/peer nodes and create a secure, reliable and large distributed storage system. The goal is for each one of them to act as if they have immediate access to a pool of information that is larger than they could hold themselves, and into which they can contribute new stuff in a both open and secure manner. Such a contributory and self-scaling community storage system is particularly useful where reliable infrastructure is not readily available in that such a system facilitates easy ad-hoc construction and easy portability. We also develop a collaborative offline Wikipedia system where the information of the whole Wikipedia is spread using our novel erasure codes among a group of community members, whose online access to Wikipedia may not be available due to some sort of failure such as a natural disaster. In another application scenario, we develop a novel framework of image sensing with a group of image sensors. The goal is to present a set of novel tools in which software, rather than humans, examines the collection of images sensed by a group of image sensors to determine what is happening in the field of view.

Participant(s)/Committee

Joseph ParadisoV. Michael Bove, Jr.Ramesh Raskar