Combining design and performance in a data visualization management system
Presented at the Biennial Conference on Innovative Data Systems Research (January 2017), Chaminade, California, USA
Interactive data visualizations have emerged as a prominent way to bring data exploration and analysis capabilities to both technical and non-technical users. Despite their ubiquity and importance across applications, multiple design- and performance-related challenges lurk beneath the visualization creation process. To meet these challenges, application designers either use visualization systems (e.g., Endeca, Tableau, and Splunk) that are tailored to domain-specific analyses, or manually design, implement, and optimize their own solutions. Unfortunately, both approaches typically slow down the creation process. In this paper, we describe the status of our progress towards an end-to-end relational approach in our data visualization management system (DVMS). We introduce DeVIL, a SQL-like language to express static as well as interactive visualizations as database views that combine user inputs modeled as event streams and database relations, and we show that DeVIL can express a range of interaction techniques across several taxonomies of interactions. We then describe how this relational lens enables a number of new functionalities and system design directions and highlight several of these directions. These include (a) the use of provenance queries to express and optimize interactions, (b) the application of concurrency control ideas to interactions, (c) a streaming framework to improve near-interactive visualizations, and (d) techniques to synthesize interactive interfaces tailored to end-users.
|Biennial Conference on Innovative Data Systems Research|
|Organisation||Centrum Wiskunde & Informatica, Amsterdam (CWI), The Netherlands|
Wu, E, Psallidas, F, Miao, Z, Zhang, H, Rettig, L, Wu, Y, & Sellam, T.H.J. (2017). Combining design and performance in a data visualization management system. In 8th Biennial Conference on Innovative Data Systems Research.