Youtopia: A Community Database Management System

This thesis introduces Youtopia, a system for collaborative management of relational data. In the age of Web 2.0, the sharing of relational data by communities is an increasingly important phenomenon. As a data management setting, it poses unique technical challenges which warrant dedicated solutions. We focus on two key aspects of Youtopia functionality: data cleanliness maintenance and handling interference between user tasks – or transactions – that access the same data. We present a set of operations for maintaining data cleanliness in a collaborative manner. A unique feature of our operation set is the mechanism for enforcing tuplegenerating dependencies (a formalism for constraint maintenance). In Youtopia, these are enforced through a process based on the classical chase that involves both automated and human-assisted steps. Next, we examine the issue of transaction interference. We provide a suitable definition of serializability for the Youtopia transaction model and present an algorithm framework that can be used to enforce it. We illustrate our framework with an extended case study that includes experimental results from the implementation of our algorithms. Finally, we begin the investigation of concurrency control notions other than serializability. We present several isolation levels for transactions which are less restrictive than full serializability, but can be implemented with much more lightweight mechanisms and thus enable the system to achieve better throughput and latency in the processing of user operations