Agroecosystem models that can incorporate management practices and quantify environmental effects are necessary to assess sustainability-associated food and bioenergy production across spatial scales. However, most agroecosystem models are designed for a plot scale. Tremendous computational capacity on simulations and datasets is needed when large scales of high-resolution spatial simulations are conducted. We used the message passing interface (MPI) parallel technique and developed a master-slave scheme for an agroecosystem model, EPIC on global food and bioenergy studies. Simulation performance was further enhanced by applying the Vampir framework. On a Linux-based supercomputer, Cray XT7 Titan, we used 2048 cores and successfully shortened the running time from days to 30. min for a global 30. years of modeling of a bioenergy crop at the resolution of half-degree (62,482 grids) with the message passing interface based EPIC (mpi_EPIC). The results illustrate that mpi_EPIC using parallel design can balance simulation workloads and facilitate large-scale, high-resolution analyses of agricultural production systems, management alternatives and environmental effects.
Bibliographical noteFunding Information:
This research was thanks to computational resources made available through the Oak Ridge Leadership Computing Facility, located in the National Center for Computational Sciences at Oak Ridge National Laboratory. Contributions were also supported by the U.S. Department of Energy (DOE) Bioenergy Technologies Office . Oak Ridge National Laboratory is managed by UT-Battelle LLC for the Department of Energy under Contract DE-AC05-00OR22725.
© 2014 Elsevier B.V.
- High performance computing (HPC)
- Message passing interface (MPI)
- Parallel design