An Extensible Self-adapting Dissemination Framework For High-speed Continuous XML Streaming Data
Okorodudu, Anthony E.
MetadataShow full item record
The efficient dissemination of data has become increasingly important with recent advances in technology that provide us immediate access to information at our fingertips. The insatiable demand for data has increased significantly with the popularity and widespread use of smart phones and tablets throughout the world. This thesis is focused on efficient push-based means of disseminating data to a large number of interested clients using a network of co-operating machines, called brokers.There have been extensive studies in optimizing both XML filtering and XPath-based subscription management in publish/subscribe systems with the aim of disseminating data in a timely fashion. Much less attention has been devoted to the optimization of the underlying broker network overlay in a dynamic environment where the subscription and publication demographics evolve over time. Although XML filtering and subscription management optimizations are important in reducing publication latency, their effects can be marginalized by a subpar broker network configuration. The effects of network configuration are exacerbated under various constraints, such as network bandwidth. Optimizing the network through manual means is impractical due to the sheer size of possible alternate network configurations. In this thesis, we present an extensible large-scale self-adapting publish/subscribe system for disseminating streaming XML data. We introduce DOXTOR (Dissemination of XML Through Optimized Routing), a distributed self-adapting publish/subscribe system. Our experimental results show that our distributed self-adapting algorithm improves the fitness of the broker network overlay over time with respect to a given cost function. We also address the issue of data loss during network reconfiguration, which has been largely overlooked in this area.