Dynamic Resource Management For IaaS Cloud

Infrastructure as a Service(IaaS)is one of the cloud computing paradigms which concentrate on the layer of infrastructure computing resources.IaaS platform provides an elastic way to require computing resources in a pay-as-you-go manner.However,IaaS cloud has some problems demanding prompt solution,one of which is how to achieve energy saving with a low Service Level Agreement(SLA)violation rate by dynamic resource management.Usually,the measure for energy saving is to improve resource utilization and thus decrease the number of active servers,while the measure for achieving a low SLA violation rate is to provide adequate computing resources for applications,so they are conflicting is a degree.Existing research works either focus on only one aspect,or do not take the overall resource utilization balancing into account,so their effects can be further improved.In this paper,we have studied in depth the resource scheduling technology,and devise an heuristic resource scheduling algorithm,and propose a hybrid optimization method for live-migration.Specifically,our work can be summarized as follows:1.We propose a novel conception,Healthy Threshold(HT),based on the traditional double-thresholds method,to indicate the healthy resource utilization status.The heuristic algorithms based on HT can regulate the overall resource utilization to be more balanced and closed to HT,so both energy saving and low SLA violation rate can be achieved effectively.2.We build an analytical model to predict the performance of live-migration.Based on the model,we propose a hybrid optimization method,which can generate different live-migration solutions according to different migration scenarios.3.We conduct large-scale simulation experiments on CloudSim to verify the effec-tiveness of our heuristic algorithms and conduct experiments on Qemu-KVM to validate that our hybrid optimization method for live-migration can improve the efficiency of resource scheduling in IaaS cloud.4.We implement a dynamic resource management system for the representative IaaS cloud-OpenStack based on our heuristic algorithms and modify the modules rela-tive to live-migration in both OpenStack and Qemu-KVM to implement the hybrid optimization method.