【作者】 王志远；

【导师】 沈纲祥；

【作者基本信息】 苏州大学 ， 信息与通信工程， 2021， 硕士

【摘要】 云计算业务的高速发展和日益普及需要大量的计算和存储资源,这导致了数据中心(Data Center,DC)的数量和规模的迅速扩张。随之而来的是数据中心的能耗也迅速增加,已成为全球电力消耗中不可忽视的一部分。为了应对数据中心的高能耗问题,业界已提出了多种用于提高数据中心资源利用率、降低能耗的方法。这其中,虚拟数据中心(Virtual Data Center,VDC)是最具代表性的一种技术方案,它能有效地提高数据中心的资源使用效率。VDC通过用户之间共享物理基础设施,可以实现计算和存储资源的高效共享。针对VDC业务,如何高效、节能地为其分配资源是一个十分重要的问题;同时,如何提高VDC业务的可靠性,从而保障在底层硬件设备或软件发生故障后,VDC业务仍不受影响,也是一个十分关键的问题。为提升数据中心的资源利用率和能效,实现绿色虚拟数据中心映射,本文将动态调频(Dynamic Frequency Scaling,DFS)技术融入到VDC映射方法中。DFS技术使得数据中心中每个设备的工作频率可以根据负载进行自适应调整,这样可以在映射VDC业务时,最大限度地降低数据中心IT设备的整体能耗。本文为映射问题构建了整数线性规划(Integer Linear Programming,ILP)模型,并提出了高效的启发式算法。通过仿真研究,发现与不采用DFS技术的VDC映射相比,在VDC映射中融入DFS技术能显著提高数据中心的资源利用率和能效,同时提出的启发式算法也很高效,可以实现与ILP优化模型相近的性能。其次,为保证VDC业务的可靠性,在单服务器故障的假设下,本文提出了两种资源保护策略:“1+1专用备份资源保护”和“共享备份资源保护”。其中,1+1专用资源保护策略为VDC业务中的每个虚拟机(Virtual Machine,VM)提前预留了专属的备份VM来进行保护,同时也预留了专用保护链路来保障VM之间的通信。但是这些备份资源在未发生故障时,都处于闲置状态,利用率十分低下。因此,为了尽量降低备份资源的开销,本文考虑在VDC业务内部共享使用VM备份资源和保护链路资源,从而在实现业务保护的前提下,尽可能地降低闲置的保护资源。本文分别针对这两种策略提出了高效的启发式算法,并进行了仿真研究。结果表明,与1+1专用备份资源保护策略相比,共享备份资源保护策略能显著地减少备份资源的使用量,并大幅度降低系统成本。更多还原

【Abstract】 The rapid growth and increasing popularity of cloud computing services requires large amounts of computing and storage resources,which has led to a rapid expansion in the number and scale of data centers(DCs).With this,the energy consumption of DCs has also increased rapidly and become a non-negligible part of global power consumption.To tackle the energy consumption issue of DCs,the industry has proposed various methods to improve DC resource utilization and reduce energy consumption.Virtual data center(VDC)is one of the most representative solutions that can effectively improve the resource utilization of DCs.For VDC services,how to allocate resources for them in an efficient and energy-saving way is a very significant issue.Also,how to improve the reliability of VDC services,such that VDC services can be fully restored when encountering the failure of underlying hardware equipment or software,is a very critical issue.To improve the resource utilization and energy efficiency of DCs and achieve green VDC embedding,we integrate the dynamic frequency scaling(DFS)technology in VDC embedding approaches.The DFS technology allows the operating frequency of each device in DCs to be adaptively adjusted according to its workload,thereby minimizing the overall energy consumption of the whole DC after embedding all VDC services.We develop an Integer Linear Programming(ILP)model for the VDC embedding problem and propose an efficient heuristic algorithm.Through extensive simulation studies,we note that integrating the DFS technology in VDC embedding can significantly improve the resource utilization and energy efficiency of DCs compared to the scheme without the DFS technology.Also,the proposed heuristic algorithm is efficient and can achieve a performance very close the ILP optimization model.Second,to enhance the reliability of VDC services,under the assumption of single-server failure,we propose two resource protection strategies:"1+1 dedicated backup resource protection" and "shared backup resource protection".The 1+1 dedicated resource protection strategy reserves dedicated backup VMs for each virtual machine(VM)in a protected VDC service in advance;it also reserves dedicated protection virtual links to guarantee the communication of the backup VMs with other VMs.However,these backup resources are often unused when there is no failure occurred.Therefore,to minimize the requirement of these backup resources,we consider sharing the use of VM backup resources and protection link resources within each VDC service,such that the protection resources can be minimized while the service protection can be guaranteed.We propose efficient heuristic algorithms for the two strategies and conduct simulation studies.The results show that the shared backup resource protection strategy can significantly reduce the backup resources required and the system cost compared with the 1+1 dedicated backup resource protection strategy.更多还原