Credibility-based Resource Scheduling Strategy Under Cloud Platform

Computing capacity in the past few years could not meet the current needs of people’sincreasing demand for computing ability, quality of service of software and demand forlarge-scale processing of data, then cloud computing is proposed. By the development of cloudcomputing, it has a very wide range of applications whether in academia or in the business world.Development of the technology makes data scale level rising from the earliest GB stage to TBstage and even PB stage. So it is imminent to considering a better platform for cloud computingservices.At the beginning of this paper the definition of cloud computing has been summed up.Mention that computing ability can provide for users as a commodity with a service fee, thenenumerated the characteristics and architecture of cloud computing system, and made a detailedanalysis of the key technology for achieving the cloud computing, at last, this paper introducedtoday’s popular cloud platforms. In order to simulating the proposed algorithm on the cloudplatform, we study the principles of MapReduce mechanism, the implementation process andHadoop’s architecture. Meanwhile, in order to compare the differences and the similaritiesbetween proposed algorithm in this paper and Hadoop resource scheduling algorithms, three oftoday’s popular job scheduling algorithms: FIFO queue scheduling algorithm, fair schedulingalgorithm and Capacity algorithm based on computing ability had been studied in this paper. Inorder to make a more detailed comparison, we analyzed the pros and cons of each algorithm.Because of the huge number of nodes resources which were distributed on the cloudplatform, this inevitably resulting in the emergence of the unreliable resource nodes. These nodeswill have a huge effect on the execution of the application and scheduling tasks. In this article,inspired by the Bayesian cognitive model and guided by the trust relationship models ofsociology, we first proposed a new cognitive trust model based on Bayesian method, and thismodel were applied to the resource scheduling system. Theoretical analysis and simulationresults show that the proposed method can effectively meet the requirements of trust on noderesource. It will cost less time to make sure the cloud computing tasks will be executed in arelatively safe node resource pool.