Research And Implementation Of Dispersion Cloud Resource Perception And Scheduling

Cloud computing is an information processing approach about large-scale computing, Itcombine scattered cloud units together to a shared resource pool using virtual resource man-agement technology、distributed computing and Internet technologies, and provide cloudservices by three manners of demand、elasticity and measurable.These cloud units are usuallydistributed across the cluster, room and data cente, when the unit changed, massive resourcesin the resource pool presents dispersion characteristics. Forming a dispersion cloud resources,that is a distributed variable with time-varying particle cloud resources. How to use cloud re-sources’s dispersion and how to perceive accurately and schedule quickly these resources andprovide cloud services by low cost and high performance way and improve user satisfactionbecome an important issue in the cloud computing area.In this paper, we study perception and scheduling problems of diffusion cloud resourcesfrom IaaS resource providers and requester (user) two angles, the main contents are:On theone hand, we proposed diffusion resource scheduling method based on mobile Agent-awarefrom the IaaS providers’ view, which maily include: firstly,we proposed resource deploymentarchitecture for IaaS layer based on mobile Agent paradigm, and gives diffusion cloudresources perceive method: Removing strategy that does not meet the conditions to one-timecompression strategy space,which based on the distribution of job completion time predictionand by setting the confidence intervals. We introduce Skyline computational thinking to ex-tract the value resources in a fine-grained way. Secondly, we study diffusion resourcescheduling method based on mobile Agent-aware: constructing multi stages iterative decisionmodel that in the form of "search-decision-execution" to capture local characteristic of jobload in a real-time and fine-grained way by perceived information from resource layer basedon mobile Agent-aware and the use of diffusion characteristics. Adaptively partitioning(merging) scheduling interval where the work load changes violently (gently) according to thesimilar characteristics of the adjacent phase of job load, and then realizing of the overallscheduling performance improvement. On the other hand we proposed the fair schedulingmethod based on dispersion characteristics which based on mobile Agent the MapReduceframework for cloud computing and the diffusion characteristics of resources as qie from use or application’ view, which maily include: firstly, we build MapReduce distributed computingframework based on Mobile Agent and define the Mobile Agent (such as Master_MA andSlaver_MA) feature set and concrete realization of the main functions of Mobile Agent. Se-condly, we study fair scheduling method based on dispersion characteristics of cloud resource.There are introducing the progress share to define the fair and share, selecting the taskaccording to user preferences, defining performance functions to start making resourcesdecisions and give specific scheduling algorithm. This scheduling method not only tradeoffthe costs and benefits, and can provide good performance and fairness in a heterogeneouscluster.In this paper, we evaluate effectiveness of the above architecture and methods on openCirrus cluster at HP Labs and explain the potential benefits of perceive scheduling method forthis article in the form of.example verified.