Scalable Framework for Agent-Based Computing
Abstract
Prototyping a framework that will support a number of agent-based algorithms in a highly distributed and heterogeneous computing environment.
Intellectual Merit
Goal is to provide a simple solution to scale a wide variety of experiments in multiple fields. Examples of possible applications could include economic/market simulations, traffic or crowd flow control, network analysis.
Broader Impact
Make significant computing resources accessible to communities of researchers who have historically not made use of HPC systems by simplifying access to scalability, while maximizing the usage of existing resources by distributing many short-duration jobs simultaneously to idle desktops, clusters, condor pools, etc.
Use of FutureGrid
For development and testing, and as one set of resources for submitting jobs which will also run on Quarry, DiaGrid and other resources.
Scale Of Use
Modest and sporadic. To test scaling it would be ideal to have access to a large number (1000+) of cores simultaneously across multiple systems, but only required for a short duration (minutes).