A coupled framework for parameter simulation and visualization.

Preeti Malakar, Vijay Natarajan, and Sathish S. Vadhiyar.
Poster at the Grace Hopper Celebration of Women in Computing in India, Bangalore, 2010


Critical climate applications like cyclone tracking require high-performance simulations to obtain real-time forecasts and high-resolution visualization by the climate scientists for subsequent timely analysis. "On-the-fly" visualization will enable the scientists to provide real-time guidance to decision makers. Such high-fidelity simulations and simultaneous visualization require large stable storage for storing the climate data and high-bandwidth networks for transferring data from the simulation to the visualization site. A combination of high simulation rate and high I/O bandwidth leads to high rate of generation of gigabytes of output data onto the disk. This gives rise to the critical problem of storage limitation for long-running climate applications. The network bandwidth between the simulation and visualization site impacts the rate at which data is moved out from the simulation site and hence determines the amount of remaining disk space available for simulation output. Eventual unavailability of storage for simulations can result in either stalling of the simulations or loss of visualization of critical climate events. In this work, we have developed a framework which adaptively uses the processor space and adjusts the frequency of output based on the application and resource dynamics.