Since the 1950’s, there have been considerable shifts in human consumption of goods, climate, and in resource/land management – more intensive farming, more reliance on fertilizers and pesticides. In sum, the United States has shifted from the smallholder family farm in favor of the industrial agriculture style of farming. These changes pose considerable challenges for maintaining high water quality. Adam Ward, Assistant Professor in the IU School of Public and Environmental Affairs, has been conducting exciting new research focused on understanding water quality and water quantity impacts of agricultural activity in the Mississippi River basin, draining more than 1.2 Million square miles of the continental US and Canada.
To do this work, Ward has been collaborating with IU’s Science Applications and Performance Tuning (SciAPT) and High Performance File Systems (HPFS) groups to optimize the Agro-IBIS code to run in parallel on IU's Big Red 2 supercomputer using an experimental extreme performance filesystem called DCRAM. This high-throughput, low-latency Lustre file system offers 35 TB of file storage, using a combination of enterprise solid state hard drive (SSD) technology, Infiniband interconnect, and eight state-of-the-art Linux servers. DCRAM, as the smaller but much faster cousin to the DC2 data capacitor, has proven capable of managing the extraordinary I/O requirements of Ward's data-intensive simulations.
This work is helping to provide both a better understanding of the aggregation of activities on water quality and to help guide public policy. Previously, a single model simulation took Ward a week or more to complete on a single processor. Now, with the experimental hardware and optimized code, more than one hundred simulations can be completed in the same amount of time. As a result, new discoveries are coming faster than ever! Ward is now able to complete ensembles of simulations to assess changes in management practices and how they impact water quality and quantity in the Midwestern U.S. This work was supported, in part, by a grant from the National Science Foundation (NSF) Grant #s EAR 1331906 and EAR 1505309, and by Intel – who provided much of the hardware for DCRAM.