Research Of A Hierarchical And Adaptive Grid Scheduling Algorithm

Grid Computing is one of the hotspots in the research of Internet, Gridis a competitive computing and resource environment,which can contain allkinds of computing resources, and change them to a environment which hasreliable, standard and economic computing capability. Grid computingemerges as a new paradigm for next-generation computing.Grid provides services for the users of resources by gathering thedispersive and free resources, this leads to a benefit relationship betweenbuyers and sellers in economic society, someone brought forward an idea tosolve the problem of the grid task scheduling and resources using through theeconomic modules, we focus on the end user's budget in the economic model.The mostly drive is basing on the trade between the user's requirement andresource in the competing market. It is benefit for them. Not only to makesure the budget and deadline. But also to solve the conflicts betweenresources providers they are all what we have to research.This thesis researches the development of the grid, and the gridscheduling modules. They are not taking in to account communication delayin the most of the modules. And the resource is noncooperation, so we designa SLU (Service Level Union)basing on communication delay, and use theauction modules, design and implement a algorithm- Research of ahierarchical and adaptive Grid Scheduling algorithm.the thesis focuses on the contents below:1. Introducing SLU, and the method to buildup a SLU, introducing thereputation of the SLU, auctioneer allocation the task basing on the allocationfactor.2. Designing and implying the algorithm, after the SLU obtaining thetask, allocating the task basing on the value of the Banzhaf, the load of thenodes are not stability, rescheduling the task basing on the other nodescommunication delay and processing ability, make sure the resourceproviders have the tasks, get the benefit, get the excellent load.3. Implying the grid scheduling algorithm in the simulation environment,analyzing the results of the algorithm, the results indicate the response time isrelative to Communication Delay, the System benefit is improving with improving the press of the time and getting the excellent load.The third chapter described the algorithm. Given a communicationdelay value, it divides the nodes into SLU basing on the value. AfterCalculating the proportion of the nodes according as processing ability andload, Calculating the Banzhaf value, the value is greater, the allocation factoris greater, using the auction modules, auctioneer collects the character of theSLU, according to the cost-time algorithm, select the appropriate SLU, in theauction, the end user give the deadline and budget, Calculating the pressureof time, the user give the allowance for the SLU, auctioneer receives thepricing, and Calculating the allocation factor, allocation factor is related withthe SLU reputation, the reputation of the SLU is decided by history records,if a task is completed successfully, the algorithm will encourage the nodeusing pride factor, and the value is bigger, but if the task is not completed intime and in quality, the algorithm will punish the node using punishmentfactor. After receive the task, the Banzhaf value is more bigger, it is first to dothe job, considering the load balance, if the load of the node is becomingbigger or exceed the average of the load, migrating the task according to theother nodes'communication delay and processing ability, for making the taskis executed in the deadline, the other nodes is voluntary to do the job, forpreventing the benefit monopolization, the budget is composed by thecommutation and the proportion of the executed task.The fourths chapter, using the Gridsim, implementing the algorithm,buildup four nodes. divided into 2 SLU, analyzing the results, it is useful toimproving the users'Qos, assure the load balance, the design of the algorithmis consistent with the purpose of that algorithm, it is proven to be reasonable.At last , It points out the shortage of the algorithm,including improvingETC-expecting model, the shortage of the auction model, in the future, wewill taking account of the dependant job.