Abstract

The rise of super computer clusters, grids, farms, and mass distributed processing are offering new ways of utilizing multiple systems and multiple localized groups of systems for achieving the computation of cooperative goals. However, methods are still being explored to efficiently use the available computational processing that comes from it for cooperative purposes.

Graph Theory has been applied to the organization and administration of dynamic networks in many approaches, but typically only in the topological organization of interconnected but physically separated systems in which paths through a network must be maintained to connect all systems. There has not yet been found an application of graph theory that has been utilized for creating self-organizing architectures of physically and locally interconnected autonomous systems.

This research introduces new graph theory and new algorithms based on graph theory for the purpose of maintaining self-organization, self-stabilization, and superstabilization properties of physically and locally interconnected distributed autonomous systems for cooperative processing. The algorithms maintain an efficient organization of nodes within a distributed system such that they are able to quickly recover from topology changes to maintain the integrity of the distributed system. Concepts are presented which utilize this resulting underlying self-organizing system to mask failures, degrade or scale performance gracefully, and build redundancy of overlying processes executing within the distributed system.

Finally, a practical application of this distributed system is offered to dynamically create, maintain, and execute component-based-system software models.