Multiagent methods and tools of management in a service-oriented distributed computing environment

Abstract

The paper discusses the issues related to ensuring the effective scaling of the job flow generated by problem-oriented distributed computing systems in heterogeneous Grid with hybrid nodes. Considers the multiagent methods and tools of the new generation, to ensure effective management of the combined job flow of scalable service-oriented software systems and load balancing for computing resources of runtime environment. Describes a high-level tools for developing services of scalable software systems. A distinctive feature of the presented methods and tools is use of elements of the economic theory (such as elements of regulation of supply and demand of resources) in the coordinated multiagent management of cluster Grid with hybrid nodes, both at the Grid level, and at the application level. The functions of problem-oriented computing environment (user application) is realized in the form of Grid-services using technology Web Services Resource Framework and patterns of interaction with local resource managers of Grid nodes. To create the agent of application is used development environment High-performance computing Service-oriented Multiagent System Framework, developed by the authors. A number of scientific services for experimental Grid are given as examples of organization of the problem-oriented computing environment (user application) using methods and tools considered in this paper. The results of their effective functioning is analyzed.

Keywords

• проблемно-ориентированные распределенные вычисления
• мультиагентное управление
• сервисы
• инструментальные средства