6 Infrastructure 6.1 Motivation

Data Nodes, Edges Display Interactive Display Visual Analogues VisualItems in ItemRegistry User Figure 6.3: The Information Visualisation Reference Model, adapted from Heer et al.[57] 6.2 State of the Art 93 a visual analytics issue that should be better tackled by all the visualisation communities. Blending different kinds of visualisations in the same application is becoming Blending different kinds of visualisations is currently difficult more frequent. Scientific visualisation and geographic visualisation need information visualisation because they manage multi-valued data with complex topologies that can be visualised using their canonical geometry. In addition, they can also be explored with more abstract visual representations to avoid geometric artefacts. For example, census data can be visualised as a coloured map but also as a multi-dimensional dataset where the longitude and latitude are two attributes among others. Clustering this data by some similarity measure will then reveal places that can be far away in space but behave similarly in term of other attributes (e.g., level of education, level of income, size of houses etc.), similarity that would not be visible on a map. On top of these visualisation systems, a user interface allows control of the overall application. User interfaces are well understood but they can be very different in styles. 3D systems use specific types of interfaces that are very different to traditional desktop interfaces. Moreover, information visualisation systems tend to deeply embed the interaction with the visualisation, offering special kinds of controls either directly inside the visualisations (e.g., range sliders on the axes of parallel coordinates) or around it but with special kinds of widgets (e.g., range sliders for performing range-queries). Interoperability can thus be described at several levels. At the data management level, at the architecture model level and at the interface level. 6.2.2 Data Management All visual analytics applications start with data that can be either statically collected or dynamically produced. Depending on the nature of the data, visual analytics applications have used various ways of managing their storage. In order of sophistication, they are: Flat files using ad-hoc formats, Structured file formats such as XML, Specialised NoSQL systems, including Cloud Storage, Standard or extended transactional databases (SQL), Workflow or dataflow systems integrating storage, distribution and data processing. We will now consider these data storage methods, paying particular attention to Data Management for visual analytics can rely on different levels of sophistication the levels of service required by visual analytics, such as: Persistence (they all provide it by definition), Typing, Distribution, Atomic transactions, Notification, Interactive performance, Computation.