Title: Evaluating I/O Characteristics and Methods for Storing Structured Scientific Data
 
Speaker: Avery Ching, Northwestern University
                              
Date/Time: Thursday, October 13, 2005, 2:00-3:00 pm

Location: Building 980, Room 95 (Sandia NM)

Brief Abstract: Many large-scale scientific simulations generate large, structured   multi-dimensional datasets.  Data is stored at various intervals on   high-performance I/O storage systems for checkpointing,   post-processing, and/or visualization.  This time for data storage   is very I/O intensive and can dominate the overall running time of the application depending on the characteristics of the I/O access pattern.  When and how to describe these I/O access patterns greatly affects performance.  In this paper, we examine the impact of various I/O parameters and methods to store scientific data using the MPI-IO interface in an optimized parallel file system.