云环境下的资源调度算法研究
发布时间：2019-01-08 10:03
【摘要】：云计算是一种新的商业计算模型和服务模式,它将计算任务分布在大量计算机构成的不同数据中心使各种应用能够根据需要获取计算能力、存储空间和信息服务。云计算数据中心利用虚拟化技术将各种软硬件资源抽象为虚拟化资源,形成虚拟化资源池,再通过资源调度技术以“按需使用,按量付费”的原则将这些资源提供给用户使用。随着现代数据中心的规模和用户数量急剧增大,如何快速高效地动态部署数据中心的这些资源成为云计算资源调度的重要问题。因此如何在保证用户服务质量,不违反服务水平协议(Service Level Agreement, SLA)的情况下提高数据中心资源的使用效率是云环境下资源调度需要研究的主要问题。 云系统的负载均衡和最小化数据中心运营成本是云环境下的资源调度面临的性能优化和成本控制的两大关键问题。针对系统负载不均衡导致的资源浪费和系统瓶颈等问题,本文提出了基于改进模拟退火的云环境下虚拟机资源的负载平衡调度算法(Simulated Annealing Load Balancing:SALB),通过最小化物理主机负载的标准差来达到系统的负载平衡。区别于传统的SA算法中随机选取初始解和邻域解的方式,本文根据系统的实时负载情况来选取合适的初始解和产生新的邻域解。利用虚拟机迁移技术,将负载过高的物理机上运行的虚拟机迁移到负载低的物理主机上,在迁移的过程中利用模拟退火的思想以一定的概率接受劣质解从而避免陷入局部最优解。在扩展后的CloudSim平台上实现了负载平衡调度算法SALB的仿真,实验结果表明SALB能够取得优于传统的模拟退火算法和轮询调度算法更好的系统负载平衡。 针对数据中心运营成本控制的问题,本文提出了基于模拟退火思想的改进遗传算法(Simulated Annealing combined Genetic Algorithm:SACGA)用于虚拟机资源分配来降低数据中心的运营成本。通过在传统遗传算法的交叉和变异过程中加入模拟退火的思想,在进化过程中以一定的概率接受劣质解,使得遗传算法能够避免过早地陷入局部最优解和早熟现象的发生。仿真结果表明SACGA能够在保证客户服务水平协议的基础上节省数据中心的操作代价,使得系统操作代价低于使用传统的遗传算法作为资源调度策略。最后总结全文并说明下一步的研究内容。
[Abstract]:Cloud computing is a new business computing model and service model. It distributes computing tasks in different data centers composed of a large number of computers so that various applications can acquire computing power, store space and information services according to their needs. Cloud computing data center abstracts all kinds of software and hardware resources into virtualized resources by using virtualization technology to form virtualized resource pool, and then provides these resources to users by the principle of "on demand, according to payment" through resource scheduling technology. With the rapid increase of the scale and the number of users in the modern data center, how to deploy these resources quickly and efficiently becomes an important issue of cloud computing resource scheduling. Therefore, how to improve the efficiency of data center resources in the case of guaranteeing the quality of service of users and not violating (Service Level Agreement, SLA) is the main problem of resource scheduling in cloud environment. Load balancing and minimizing the operating cost of data center are the two key problems of resource scheduling in cloud environment, such as performance optimization and cost control. Aiming at the problem of resource waste and system bottleneck caused by system load imbalance, this paper proposes a load balancing scheduling algorithm (Simulated Annealing Load Balancing:SALB) based on improved simulated annealing for virtual machine resources in cloud environment. The system load balance is achieved by minimizing the standard deviation of the physical host load. Different from the traditional SA algorithm in which the initial solution and the neighborhood solution are randomly selected, this paper selects the appropriate initial solution and produces a new neighborhood solution according to the real-time load of the system. Using the technology of virtual machine migration, the virtual machine running on the overloaded physical machine is migrated to the low-load physical host. In the process of migration, the idea of simulated annealing is used to accept the inferior solution with a certain probability so as to avoid falling into the local optimal solution. The simulation of load balancing scheduling algorithm SALB is implemented on the extended CloudSim platform. The experimental results show that SALB can achieve better load balancing than the traditional simulated annealing algorithm and polling scheduling algorithm. In this paper, an improved genetic algorithm (Simulated Annealing combined Genetic Algorithm:SACGA) based on simulated annealing (SA) is proposed to reduce the operating cost of the data center by allocating virtual machine resources in order to control the operating cost of the data center. By adding the idea of simulated annealing in the process of crossover and mutation of traditional genetic algorithm, we can accept the inferior solution with a certain probability in the evolution process, so that the genetic algorithm can avoid falling into the local optimal solution and premature phenomenon prematurely. The simulation results show that SACGA can save the operation cost of the data center on the basis of guaranteeing the customer service level protocol, which makes the operating cost of the system lower than that of using the traditional genetic algorithm as the resource scheduling strategy. Finally, the paper summarizes the full text and explains the next research content.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.01
本文编号：2404453
[Abstract]:Cloud computing is a new business computing model and service model. It distributes computing tasks in different data centers composed of a large number of computers so that various applications can acquire computing power, store space and information services according to their needs. Cloud computing data center abstracts all kinds of software and hardware resources into virtualized resources by using virtualization technology to form virtualized resource pool, and then provides these resources to users by the principle of "on demand, according to payment" through resource scheduling technology. With the rapid increase of the scale and the number of users in the modern data center, how to deploy these resources quickly and efficiently becomes an important issue of cloud computing resource scheduling. Therefore, how to improve the efficiency of data center resources in the case of guaranteeing the quality of service of users and not violating (Service Level Agreement, SLA) is the main problem of resource scheduling in cloud environment. Load balancing and minimizing the operating cost of data center are the two key problems of resource scheduling in cloud environment, such as performance optimization and cost control. Aiming at the problem of resource waste and system bottleneck caused by system load imbalance, this paper proposes a load balancing scheduling algorithm (Simulated Annealing Load Balancing:SALB) based on improved simulated annealing for virtual machine resources in cloud environment. The system load balance is achieved by minimizing the standard deviation of the physical host load. Different from the traditional SA algorithm in which the initial solution and the neighborhood solution are randomly selected, this paper selects the appropriate initial solution and produces a new neighborhood solution according to the real-time load of the system. Using the technology of virtual machine migration, the virtual machine running on the overloaded physical machine is migrated to the low-load physical host. In the process of migration, the idea of simulated annealing is used to accept the inferior solution with a certain probability so as to avoid falling into the local optimal solution. The simulation of load balancing scheduling algorithm SALB is implemented on the extended CloudSim platform. The experimental results show that SALB can achieve better load balancing than the traditional simulated annealing algorithm and polling scheduling algorithm. In this paper, an improved genetic algorithm (Simulated Annealing combined Genetic Algorithm:SACGA) based on simulated annealing (SA) is proposed to reduce the operating cost of the data center by allocating virtual machine resources in order to control the operating cost of the data center. By adding the idea of simulated annealing in the process of crossover and mutation of traditional genetic algorithm, we can accept the inferior solution with a certain probability in the evolution process, so that the genetic algorithm can avoid falling into the local optimal solution and premature phenomenon prematurely. The simulation results show that SACGA can save the operation cost of the data center on the basis of guaranteeing the customer service level protocol, which makes the operating cost of the system lower than that of using the traditional genetic algorithm as the resource scheduling strategy. Finally, the paper summarizes the full text and explains the next research content.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.01
【参考文献】
相关期刊论文 前5条
1 张敏;陈云海;;虚拟化技术在新一代云计算数据中心的应用研究[J];广东通信技术;2009年05期
2 华夏渝;郑骏;胡文心;;基于云计算环境的蚁群优化计算资源分配算法[J];华东师范大学学报(自然科学版);2010年01期
3 高尚;模拟退火算法中的退火策略研究[J];航空计算技术;2002年04期
4 米海波;王怀民;尹刚;史殿习;周扬帆;袁霖;;一种面向虚拟化数字中心资源按需重配置方法[J];软件学报;2011年09期
5 高宏卿;邢颖;;基于经济学的云资源管理模型研究[J];计算机工程与设计;2010年19期
，本文编号：2404453
本文链接：https://www.wllwen.com/guanlilunwen/ydhl/2404453.html
