Management, Multi-economic Intelligent Agent-based Computing Grid Resources

Computational grid is a development direction of distributed and parallel processing technology. Resource management is one of the key techniques of computational grid. It has magnitude theoretical significance and applied value to develop computational grid.Resource management of computational gird is a challenging question because that resource in computational grid is distributed, is autonomous, is heterogeneous, is dynamic, and is need of coordination. This paper goes into the researching status of this question and points out meagerness of applying various resource management techniques in using now into computational grid.Firstly, this thesis discusses the rationality of using economic method and agent technology to manager grid resource, and then initiates Grid Resource Allocation and Management architecture based on Multi-Economic-Agents. Compared with other works, GRAMMEA not only has economic basic in theory, but also is scalable and adaptive in practice. At the same time, it supports quality of service in a certain extent.Secondly, this thesis founds the economic model of GRAMMEA. It defines the equilibrium state of resource allocation and that of service allocation, and proves the theorems about optimization fairness and existence of the equilibrium state, which are the theoretical basic of GRAMMEA architecture and assure that GRAMMEA can manager grid resource fairly and efficiently.Thirdly, this thesis constructs three trading protocols, which are the resource trading protocol, the service trading protocol and the information trading protocol. These trading protocols define the interact rules of economic agents. Every trading protocol constitutes two parts: one is the trading protocol language and the other is the trading protocol process. The former specifies the grammatical rules of trading sentence, and the later specifies the pragmatics rules of trading sentence.At last, this thesis puts forward two groups of algorithms. The first group of algorithms is about resource and service allocation, which makesallocation scheme closed to the equilibrium state in the economic model. The second group of algorithms is able to guarantee quality of service in a certain extent by using workload-oriented admission control , workload-oriented task loading, task group scheduling based on virtual clock and so on.In conclusion, based on economic models GRAMMEA with three trading protocols and two groups of algorithms provides an effective approach to manager grid resource.