Font Size: 

The Hohokam Water Management Simulation II is a reimplementation of a desktop-based simulation of the operation of a large-scale irrigation system for a high-performance computing (HPC) environment. The focus of the simulation is the examination of the Hohokam, who lived along the Salt and Gila Rivers in the US Southwest and who constructed and maintained a large-scale irrigation system that was in use for nearly a millennium. The new simulation implementation is specifically designed for the Blue Gene/P supercomputer (BG/P) maintained by Argonne National Laboratory (ANL), which can perform operations on more than 163,000 processors simultaneously. The simulation framework used is based in part on the recently-released Repast HPC platform, also being developed at ANL. The simulation makes use of the highly parallel processing environment of the BG/P in two ways: first, by executing single simulation runs across multiple processes, making possible both simulations at larger spatial scales and those at finer chronological and spatial resolution; second, by permitting large numbers of concurrent runs to be executed strategically to explore the extremely large parameter space made possible by various combinations of input data, parameters, spatial and temporal scales, and algorithms available to the simulation. A key component is the use of a relational database to maintain the connection between input and output, and to structure the variant simulations that can be run. Technical and theoretical implications of large-scale archaeological simulation modeling will be discussed, with a special emphasis on the way that the approach allows investigation into the computational aspects of the social system under study. Initial simulation results will be used to illustrate these issues and the additional challenges and opportunities of the HPC approach.

High-Performance Computing; Irrigation; Simulation; Agent-Based Modeling