Computational & Technology Resources
an online resource for computational,
engineering & technology publications
Computational Science, Engineering & Technology Series
TRENDS IN PARALLEL, DISTRIBUTED, GRID AND CLOUD COMPUTING FOR ENGINEERING
Edited by: P. Iványi, B.H.V. Topping
The Virtual Prairie Project
M. Garbey1, M. Smaoui1, W. Rinsurongkawong1 and C. Mony2
1Department of Computer Science, University of Houston, Texas, United States of America
M. Garbey, M. Smaoui, W. Rinsurongkawong, C. Mony, "The Virtual Prairie Project", in P. Iványi, B.H.V. Topping, (Editors), "Trends in Parallel, Distributed, Grid and Cloud Computing for Engineering", Saxe-Coburg Publications, Stirlingshire, UK, Chapter 3, pp 49-82, 2011. doi:10.4203/csets.27.3
Keywords: clonal plants, multiscale model, agent-based model, volunteer computing, genetic algorithm, data mining.
In this chapter, we introduce a computational framework for a multidisciplinary project, called Virtual Prairie (ViP), designed to better understand clonal plant dynamics. ViP involves a team of ecologists, mathematicians and computer scientists.
Individual based models have been the most popular approach in ecological modeling since the 1980s, but no generic model exists for multispecies clonal plant communities. We constructed a three-scale model framework  that simulates a virtual prairie: starting from the individual plant, going to the prairie and finally incorporating external abiotic factors. Such amodel can address a variety of species, habitats and environments, just by varying a set of parameters. The downside is that we ended up with dozens of parameters. Thus, taking advantage of our model was impossible without advanced parallel computing.
The ViP framework relies on a high performance computing and volunteer computing (VC). VC is an arrangement in which people provide computing resources to projects, which use the resources to do distributed computing. The Virtual Prairie BOINC project is the volunteer computing setting we have been using to perform campaigns of computations  aimed at answering multiple ecology questions. These campaigns consisted of browsing the parameter space of the model and computing multiple objective functions of the simulated prairies.
Campaigns of computations performed on the Virtual Prairie BOINC project can last up to the order of a month, while producing a huge amount of data. We needed a computer tool that can analyze quickly the results of those campaigns while they were running, in order to decide if the ongoing computations were promising or probably not relevant to our objectives. Our tool incorporates standard data mining techniques  including clustering and association rules into a web based application running on the server.
When the number of parameters increases, it is more practical to use evolutionary algorithms such as the genetic algorithm to explore the parameter space or to find the optimum of a certain objective function. The advantage of such algorithms is that they have built-in embarrassing parallelism at the level of fitness evaluation which can be done with VC. However, as a result of the nature of VC such algorithms need some adaptation to the platform. We managed to run a genetic algorithm using the Virtual Prairie BOINC project to run both benchmark and prairie optimizations. The computational framework basically relies on volunteer computing to achieve large scale simulations and perform optimizations of virtual prairies.
purchase the full-text of this chapter (price £20)