Research On Multi-layer Nested Virtualization In Optimizing Cloud Resources Usage

Cloud computing has realized people's dream of making computing resources more attractive as a service and effectively solved many problems for customers,such as resource constrain and high overhead of maintaining large-scale software and hardware.However,under the restriction of resource isolation in cloud computing,the idle resource slices of one rented cloud server can not be reused by another tenant until its lease expires.With the increasing number of the rented cloud servers,this can lead to great loss of cloud computing resources.How to efficiently reuse the idle slices has been a challenge for cloud computing.Among the existed technologies,virtualization performs brilliantly in improving the use ratio of computing resources.However,cloud servers are all virtual machines and the traditional virtualization technology can be helpless in this situation.Fortunately,nested virtualization was proposed to solve the problem.Nested virtualization allows one hypervisor to run multiple hypervisors with the associated virtual machines.However,the existing architectures can not make the greatest advantage of nested virtualization in computing resources saving.In this thesis,we research on the performances of the existing nested virtualization architectures and then propose an optimized scheme for improving the use ratio of cloud resources under the IaaS cloud service.Our contributions can be summarized as follows:(1)We propose a new scheme for optimizing cloud resources usage using multi-layer nested virtulization.By modifying the traditional nested virtualization architectures,we propose an optimized multi-layer nested virtualization architecture,which can efficiently reduce the resource slices and improve the utilization of cloud resources;Based on load balancing concept,we propose a new task scheduling algorithm for nested-cloud.Benefiting from the scheduling strategy customization function,our task scheduling algorithm can both improve the utilization of cloud resources and achieve a tradeoff between computing efficiency and service monetary costs.The main result of this thesis was accepted by International Journal of Embedded Systems,which is indexed by EI compendex.(2)We use OpenStack program to construct a private cloud platform in Linux environment and implement the performance analysis of multi-layer nested virtualization architecture on it.Besides,we also do numeral simulations of our task scheduling algorithm.The result shows that our scheme is well-designed and can effectively improve the use ratio of cloud computing resources when the extra overhead of nested virtual machine is within reasonable limits.Moreover,compared with traditional heuristic scheduling algorithm and loop distribution algorithm,our designed task scheduling algorithm can get 18% and 10%service monetary costs saving respectively.