Beyond the visualization pipeline: The visualization cascade

The concept of a pipeline has become a quite common way of thinking about the process of visualizing data. In this article we discuss the inherent limits of this concept and argue for the need to expand this concept for achieving higher performance and convenience to the end user. While the traditional model of a visualization pipeline describes the execution of some data ow, it is most suitable for a static data-set. However for time-dependent data (e.g.) we intend visualizations to be as fast in time as they are in space. The pipeline model does not recognize similarity and repetition of operations which is essential to achieve the desired performance. The pipeline model thus needs to be extended to eciently cover multiple traversals and caching of intermediate results, which we call the Visualization Cascade. It will be demonstrated in practice within its implementation in the VISH visualization environment.