Researchers in the University of Michigan Electrical Engineering Computer Science Department have developed software that provides a computational infrastructure for managing large-scale simulation studies on computing clusters. This is especially useful for agent-based modeling and analysis in finance, defense, traffic flow and many other areas. The use of cluster (or parallel) computing presents a challenge to many users, as programming time and knowledge are required to efficiently transmit, monitor, collect, and manipulate data into and within the computing clusters. The developed software provides an easy to use interface for efficient cluster computing simulations written in any programming language using any underlying simulation methodology. The software is able to run multiple simulation experiments simultaneously, re-use data, and iteratively refine experiments.
Businesses and researchers are performing more large-scale simulations as computer models become more prevalent and cluster computer becomes more accessible. These large-scale simulations, including multi-agent based simulation (MABS), provide modeling and analysis to a wide range of applications such as algorithmic trading, supply chain analysis, product adoption, stem cell development, traffic flow, and a host of other areas. The developed software addresses the challenge of providing easy access to computer clusters by combining numerous factors to provide efficient scheduling and robust data storage solutions. The software has been tested over a variety of simulator programs, processing over 8 million observations. The software provides an easy gateway between the designer of the simulation and the hardware it is run on, providing a faster, easier way to obtain the larger, more powerful simulations.
- Defense simulations
- Traffic flow analysis
- Supply chain analysis
- Algorithmic trading optimization
- Faster to implement simulations
- Easier access to commercial computing clusters
- Supports any programming language and simulation