Visualizing time-varying call graphs is challenging due to vast amounts of data at many dimensions to be displayed: Hierarchically organized vertices with attributes, directed or undirected edges with weights, and time. In this paper, we introduce a novel overview representation that shows the dynamic graph data as a timeline- and pixel-based aggregated view targeting the preservation of a viewer's mental map by encoding the time-varying data into a static diagram. This allows us to easily and effortlessly compare dynamic call graphs on different levels of hierarchical granularity. The proposed data extraction and visualization system uses this overview as a good starting point for further and more detailed investigations by applying existing dynamic graph visualization techniques that show the graph structures and properties more clearly. These more task-specific visualizations show the dynamic graph data from different perspectives such as curved node-link diagrams or glyph-based representations combined by linking and brushing. Intermediate analysis steps can be stored and rebuilt at any time by using corresponding thumbnail representations.
This paper will be presented at the VMV workshop that is held in November 2012 in Magdeburg, Germany. I would like to thank the 12 software engineering students for this great work. If you have any questions about the project please contact me.