Query load balancing for incremental visible object extraction

Interactive visual data explorations impose rigid real-time requirements on the extraction of visible objects. Often these requirements are met by deploying powerful hardware that maintains the entire data set in huge main memory structures. In This work we propose an approach that retrieves the visible data on demand and is based on a tight integration of the database and visualization systems. We propose to incrementally adjust the observer path by adding and dropping path points. The result is an optimal path that minimizes the interaction with the database system and retrieves all visible objects along the path. The visible objects are retrieved incrementally, and it is possible to precisely control the query load and the number of retrieved objects. The minimal distance path method issues frequent queries and retrieves the lowest possible number of objects at each query point. The end point path method minimizes the number of queries by permitting larger query results. The experimental results show that our implementation reduces the number of I/O operations significantly, is stable, and does not depend on the initial number of path points.