Conceptual Model of Problem-oriented Heterogeneous Distributed Computing Environment with Multi-agent Management

Abstract

One of the promising approaches to the efficient management in a heterogeneous distributed computing environment is the use of multi-agent systems. When researchers want to solve tasks in such an environment, they need utilities for the description of the processing in the specific domain, and knowledge about a software-hardware infrastructure. This multi-component knowledge is used for effective decision-making. Practice and experience of multi-agent systems show that knowledge representation is non-trivial problems. We proposed an approach to the design of the conceptual model of a heterogeneous distributed computing environment based on multi-agent management. An important component of the model is algorithmic knowledge, which provides the support of planning, resources allocation and dynamic decomposition of a problem to subproblems. This knowledge is represented as a semantic network, which describe semantic relations between concepts. Unlike to Semantic Web Services and Semantic Grids, our approach provides advanced planning abilities to support parallelism and scalability. We develop specialized high-level tools for designing and using the conceptual models.