We are providing simulation tools, in a high performance setting, for Enterprise Modeling, Social Simulation war gaming, and related computations. The goal of this work is to produce a framework that hides the complexities of parallel communication to allow application developers to deal directly with the task at hand.
The first instantiation of this work is the Parallel Particle Data Model (PPDM). The PPDM allows a developer to break up their problem domain into units called cells and disperse populations of particles across these cells. These populations can represent things like people, cars, or even abstract idea such as a social interaction.
In the background the cells and particles are distributed across multiple processors. The developer submits Particle Algorithms to the system which are also distributed across the processors to interact with these particles. These Particle Algorithms are given access to the particles on a cell by cell basis and they are only ableto see cells owned by their host processor as well as subsets of ghost cells found on neighboring processors. As particles are updated and moved from cell to cell the framework takes care of all of the necessary communication. This framework will allow developers to attack problems spaces that had previously been too computationally or memory intensive to attempt before.