Grid Environment Task Division And Utility Maximization Model Of The Allocation Of Resources

As the computer science and application technology, and networks technology hasdevelopped rapidly in recent years, grid computing emerges. It is a distributed andheterogeneous computing environment, characteristed by its large-scale users,heterogeneous and scatteredly distributed resources, dynamic and demand diversity etc.Users share the network resourses via making a request to the system in grid, while thesystem charges from users according to their service function and performance demands.Consequently, in order to maximize the provider’s profit as well as to meet the users’performance needs for service efficirncy, service reliability or data security, the gridsystem should be capable of seeking the best scheduling strategy among numerousfeasible methods intelligently.Considered the above, this thesis presents a task division and resource allocationstrategies optimization approach in grid computing，in order to obtain maximal serviceprofits under different conditions. And the general generation function and geneticalgorithm are used. Besides, illustrative examples are given for test and validation.First, based on the the previous work, the task division and resource allocationrealization mechanism in grid system is investigated and gradually built considering thecharacteristics. When received a service request from users, the grid system analyzes thetask’ characteristics and divides it into a set of small and low-coupling execution blocksthat can be executed in parallel. Then, suitable resources are chose and further assignedto the current task for exercution, after the analysis on the available resources’application characteristics parameters. After the resources finish the assigned jobs, theyreturn the results back to the system and then the RMS integrates the received resultsinto entire task output which is requested by the user.Second, the various service quality requirements of different individuals are minedfrom the views of the service users and service providers respectively. What’s further,several performance evaluation indexes to the grid system and their mathematical modelare proposed, including service efficirncy, service reliability, data security and servicecost and profits. And then the general generating function technology is adopted to calculate each index quickly for the service profit maximization model.At last, genetic algorithm is used to solve the proposed task division and resourceallocation scheduling optimization problems in grid environment. Illustrative examplesare also designed for testing the service profit maximization model under differentconditions and the optimal solutions are discussion from different perspectives.