Multiagent Approach for Building Distributed Adaptive Computing System

AbstractThe new method of the organization of the distributed adaptive computations is offered in the presented paper. The method allows using resources of computers united by a network for solving coherent tasks. The main feature of the given method is adaptive correcting of process of computation if parameters of computing nodes are varying or nodes fail. This ability is achieved because of system decentralization and multiagent approach: all nodes of computing system are equal and nodes have to unite to solve incoming tasks. Also this paper presents algorithms based on the developed method and program model of decentralized computing system