Event processing applied to streams of TV channel zaps and sensor middleware with virtualization
MetadataVis full innførsel
OriginalversjonEvent processing applied to streams of TV channel zaps and sensor middleware with virtualization by Pål Evensen, Stavanger : University of Stavanger, 2013 (PhD thesis UiS, no. 192)
The last decade has seen an exponential increase in mobile computing devices, as well as an increasing adoption of sensor technology in process industry, homes and public spaces. The increasing amount of information made available by such devices has led to a class of pervasive systems that require little or no user input. Smart home systems is an example of such pervasive systems. A main obstacle for application developers dealing with sensor-based systems is heterogeneity of devices and protocols. A common obstacle for end-users is the manual configuration of networked devices. Our first research contribution is a middleware that overcomes these obstacles: The SENSEWRAP middleware addresses the problem of heterogeneity in a smart home setting through the virtualization of hardware and services. Furthermore, it provides automatic network configuration and service discovery. The usefulness of pervasive systems usually correlates with their ability to perform their functions in the background, without user involvement. Instead, these systems base their actions on available information relevant to their application, e.g., they are information-driven. For information-driven systems, like smart-home systems and other pervasive systems to be able to decide on the correct action at the right time, it is vital that the correct information is made available to them in a timely manner. A primary asset of publish/subscribe interactions is the immediate distribution of new information available to interested parties, and as such, it is a well-suited model for building highly scalable and flexible systems that are able to cope with a dynamic environment. Complex event processing is a fairly new paradigm that refers to the processing and correlation of events as they occur. There exists several specialized programming languages for performing complex event processing. A main goal of such languages is to enable the programmer to express patterns of events in a simpler and more straightforward manner than what is possible with a generalpurpose programming language. A main contribution of this thesis is an exploration of the tradeoffs involved in using a specialized, declarative event processing language versus using a general-purpose, imperative programming language for event processing applications. Our results indicate that going the specialized language route does indeed simplify development of event processing applications, but that this comes at the expense of performance. Furthermore, we present the EVENTCASTER platform for building eventbased systems, on which we have built two novel event processing applications: The viewer statistics and ADSCORER applications are research contributions in their own right. The viewer statistics application demonstrates how event processing techniques can be applied to broadcast television, in order to provide more accurate viewer statistics than what is currently available, in near-real time. With the ADSCORER application, advertisers and broadcasters are provided with a detailed evaluation of each individual advertisement, previously only available to advertisements distributed on the web.
PhD thesis in Information technology