Researches On Key Technologies In Grid Resource Management

Grid is to use Internet technology to connect, assign, manage and coordinate the different resources which are geographically dispersed in different locations, including computing resource, storage resource, communication resource, software resource, information resource and knowledge resource etc. And all the resources compose a virtual supercomputer via the logic relations. It makes the effective management and scheduling of Grid resource more complex than traditional distributed network due to the distribution, heterogeneity, dynamic, autonomy and other characteristics of Grid resource. Consequently the Grid resource management becomes an important research domain in Grid research.Based on a systematical summary of relevant works on Grid resource management, this dissertation focuses on four key technologies in Grid resource management. The major contributions of this dissertation are as follows:(1) This dissertation proposes a Grid resource management model based on mobile agent. It utilizes the autonomy and intelligence of mobile agent to manage the distributed, heterogeneous, autonomous and dynamic Grid resource. It is a new management model and a feasible management method.(2) This dissertation presents a hierarchy Grid resource organization model based on semantics. There exists poor expansibility, overload bandwidth etc. in current Grid resource organization model. To solve these problems, this dissertation inducts the semantic tree organization manner to organize the Grid resource based on the Grid resurce clustering result in the mobile agent-based Grid resource management model. The model provides a good organization infrastructure to the Grid resource search. The dissertation validates the validity of the model via experiment.(3) This dissertation proposes a Grid resource search strategy based on ontology. There exists poor reliability and low efficiency in current Grid resource search methods. The strategy inducts ontology to complete the Grid resource search in the mobile agent-based Grid resource management model to solve these problems. And it discards the traditional symmetrical request-resource description in Grid resource search and utilizes the asymmetrical request-resource description manner. It can construct the resources ontology, request ontology and strategy ontology based on a set of proposed rules in the SemanticObjects platform. The resource and users in the Grid can present their information according to the obtained ontology to optimize the match between the resources and request. The strategy can resolve the existing problems in current Grid resource search methods well.(4) This dissertation presents a Grid resource schedule strategy based on an improved coarse grain genetic algorithm in the mobile agent-based Grid resource management model. There exists single point of failure, high cost of communication and low efficiency of schedule algorithm in current Grid resource schedule. To solve these problems, the strategy inducts the basic theory of genetic algorithm and utilizes its distributed and parallel characters to solve the strong distribution and parallel in Grid resource scheduling. Due to the strong robust, low requirement for the initial condition and no intervene in running process of the genetic algorithm, it can control the Grid resource scheduling well. Farther the strategy utilizes the improved coarse grain genetic algorithm to enhance the scheduling efficiency.