hoc Computational Clusters of Mobile Devices 3 Raw data Prepared Data Focus Data Geometric Data Image Data Data Analysis Filtering Mapping Rendering

Current web-based visualizations are designed for single computers and cannot make use of additional devices on the client side, even if today’s users often have access to several, such as a tablet, a smartphone, and a smartwatch. We present a framework for ad-hoc computational clusters that leverage these local devices for visualization computations. Furthermore, we present an instantiating JavaScript toolkit called VISHIVE for constructing web-based visualization applications that can transparently connect multiple devices—called cells—into such ad-hoc clusters—called a hive— for local computation. Hives are formed either using a matchmaking service or through manual configuration. Cells are organized into a master-slave architecture, where the master provides the visual interface to the user and controls the slaves, and the slaves perform computation. VisHive is built entirely using current web technologies, runs in the native browser of each cell, and requires no specific software to be downloaded on the involved devices. We demonstrate VisHive using four distributed examples: a text analytics visualization, a database query for exploratory visualization, a DBSCAN clustering running on multiple nodes, and a Principal Component Analysis implementation.