Efficient input and output for scientific simulations
Sy‐Yen Kuo, Marianne Winslett, Y. Cho, J. Lee, Y. Chen · 1999
Large simulations which run for hundreds of hours on parallel cornputers often periodically generate snapshots of states, which are later post-processed to visualize the simulated physical phenomenon.For many applications, fast I/O during past-processing, which is dependent on a" efficient orgaoization of date on disk, is as important as minimizing computation-time I/O.In this paper we propose optimizatioo.5 to support efficient parallel I/O for scientific simulalions and subsequent visualizations.We present an ordering mechanism to linearize date on disk, a performance model to help to choose a proper stripe onit size, and a scheduling algorithm to minimize commonication contention.Our experiments on a" IBM SP show that the combination of these strategies provides a 202.5% performance boost.