Storz, Oliver Dirk and Davies, Nigel (2008) A distributed systems infrastructure for open public display research networks. PhD thesis, Lancaster University.
Storz2008_Thesis.pdf - Published Version
Available under License Creative Commons Attribution-NoDerivs.
Download (9MB)
Abstract
Fuelled by a sharp decline in the cost of large-screen display hardware, we are currently witnessing a continuous increase in the number of public displays being deployed in locations, such as airports, railway stations and shopping malls. These commercial deployments range from single displays in shop windows, to hundreds of displays in airports and train stations. In parallel, public display systems have provided researchers with a rich topic for research in a range of disciplines, including HCI, CSCW and sociology. However, once deployed, public displays represent a publicly visible “face” of the corporation or entity behind the deployment and the introduction of experimental, research-led content is therefore often undesirable. With advertising being a driving factor behind the deployment of many displays, stake holders are also less likely to give up valuable “airtime” to experimentation. Researchers willing to overcome these issues by deploying their own hardware, e.g. in a lab or campus environment, discover that the software packages used to drive commercial displays are tailored towards the needs of commercial deployments and are unable to meet the requirements for experimentation and research. As a result, researchers using public displays as a vehicle for their investigations normally create small-scale, closed deployments in lab or office environments aimed at investigating specific research questions. Applications and content are usually hand-crafted to best support the investigations at hand, and deployments are rather short-lived. While other research communities, e.g. in the context of PlanetLab or the Grid, have identified the benefits of creating open, shared, medium or large-scale research infrastructures, a similar move has so far not taken place in the context of public display networks. This thesis aims to provide the foundations for the creation of medium-scale, long-lived public display networks for experimentation and research. The thesis presents a distributed systems infrastructure and API for providing experimenters with the means for creating and running interactive experiments on public display networks. Moreover, it introduces a computational model that provides key abstractions over the hardware and software entities present in a public display network. The provided API is split into two parts: a low-level API, enabling experimenters to directly control the life-cycle and visibility of experimental content on displays, and a high-level API and scheduling service for scheduling experiments based on constraints, such as time, location, and the presence of Bluetooth devices (e.g. mobile phones). The low-level API features transactional semantics that provide atomicity, consistency and visual isolation for orchestrating experiments involving multiple displays. The constraint-based scheduling interface of the high-level API builds on and complements the low-level API, and provides experimenters with an easy-to-use interface for running and displaying content. The software infrastructure presented in this thesis has been implemented and deployed as part of the e-Campus public display network at Lancaster University. The software infrastructure is used on a daily basis for scheduling a mixture of research-led experimental content and informa- tional content. We report on both a quantitative and qualitative evaluation of the system based on performance measurements and a study of longitudinal use.