(455b) Migrating High Performance Computing to the Amazon Cloud - Methods and Examples Using Environmental Systems Analysis

To adequately analyze a subsurface environmental system, including assessing the inherent uncertainty in the information available to characterize it, and the ability to robustly include the governing physical processes requires numerical computational models. Since these models require inclusion of complex biogeochemical processes and simulate large aerial extents, each model run can require execution times on the order of hours or days to complete.

Adequate analysis of the uncertainty inherent in these subsurface systems performances in their natural state, or due to imposed stresses such as experienced during active dewatering or groundwater remediation, can require 100s or 10,000s of simulations be conducted. Thus, an analysis of complicated subsurface systems can require 100,000s of hours of computer processing time.

With the advent of Cloud computing, computation time on an as-needed basis is readily available. This work discusses implementing large-scale computational analysis on the Amazon Cloud including determining a capacity plan, configuring the project workload, managing the grid engine worker queue, staging and storing results for post processing. Three examples of various size and complexity of actual Cloud computation projects are provided with experience and lessons learned.