Streaming data management for the online processing of simulation data

The fast processing (storing; computing control variates) of simulation data is essential for efficient simulation. The impact of poor data processing might annihilate all the efforts invested into designing high-performance algorithms for computing the trajectories. Streaming data management can be used as an alternative to the more classical approaches (in-memory and write-and-read-back mechanisms). In first experiments our implementation of streaming data management shows better run time performance than in-memory solutions, although the streaming additionally integrates online calculations, e.g., to determine variance and average.