Adaptive and Efficient Transfer for Online Remote Visualization of Critical Weather Applications

Critical weather applications such as cyclone tracking require online visualization simultaneously performed with the simulations so that the scientists can provide real-time guidance to decision makers. However, resource constraints such as slow networks can hinder online remote visualization. In this work, we have developed an adaptive framework for efficient online remote visualization of critical weather applications. We present three algorithms, namely, most-recent, auto-clustering and adaptive, for reducing lag between the simulation and visualization times. Using experiments with different network configurations, we find that the adaptive algorithm strikes a good balance in providing reduced lags and visualizing most representative frames, with up to 72% smaller lag than auto-clustering, and 37% more representative than most-recent for slow networks.