云计算环境下资源调度关键技术研究

发布时间：2018-06-02 02:09

本文选题：云计算 + 云数据中心　；参考：《北京邮电大学》2014年硕士论文

【摘要】：云计算是当下具有巨大潜力价值的新兴计算技术,它利用大规模的硬件和虚拟资源为用户提供动态的应用服务。为了最大程度的利用云资源,充分发挥云计算的最大潜能,挖掘高效的资源调度策略是我们的当务之急。资源调度策略负责在异构云资源池中选择匹配合适的资源来执行对任务请求的处理。在制订资源调度策略时应充分考虑云计算的实际场景,以寻找到最优的调度方案。本文通过学习云计算资源调度相关技术,深入对现有调度算法进行研究分析。然后主要针对两种场景下的云资源调度策略进行优化,旨在提高服务质量的同时提高系统性能。本文的主要研究工作如下：首先,云环境下的应用存在一定数量的轻量级任务。将细粒度任务请求配置到高计算性能的资源池会增加系统整体的等待时间和周转时间。大量的细粒度任务将会花费很多时间在任务调度和传输上。此外,细粒度任务分配至高计算性能的资源节点会明显的降低资源利用率。针对该云场景,基于资源优先级的动态资源调度策略并将细粒度任务以分组方式整合后再执行处理。同时综合考虑资源的带宽状况,对调度策略加以优化。其次,当前的云资源调度策略在资源节点的负载均衡上并不理想,易出现节点间的负载不均衡。现存的调度算法大多数并未考虑用户群体的差异,致使VIP用户并不能获得更优质的服务。为了解决上述云系统瓶颈,以Min-Min调度算法作为基础进行研究分析。该算法的复杂度较低易实现,但其短板在于资源负载不均。因此针对此现象,首先对经典Min-Min调度算法加以进化,使其负载均匀分布并保证VIP级服务质量。最后,对论文的研究内容进行总结陈述。整理陈述本文的研究成果,并以对资源调度技术的系统学习作为基础,展望未来的研究方向。
[Abstract]:Cloud computing is a new computing technology with great potential value at present. It uses large-scale hardware and virtual resources to provide users with dynamic application services. In order to maximize the use of cloud resources, full play the maximum potential of cloud computing, mining efficient resource scheduling strategy is our urgent task. Resource scheduling strategy is responsible. Select the appropriate resources in the heterogeneous cloud resource pool to perform the task request processing. In making the resource scheduling strategy, the actual scene of the cloud computing should be fully considered in order to find the optimal scheduling scheme.
In this paper, we study and analyze the existing scheduling algorithms through learning the technology of cloud computing resource scheduling. Then we mainly optimize the cloud resource scheduling strategy under the two scenarios, aiming at improving the quality of service and improving the performance of the system.
First, there are a certain number of lightweight tasks for applications in the cloud environment. Configuring fine grained task requests to a high computing resource pool will increase the overall waiting time and turnover time of the system. A large number of fine-grained tasks will spend a lot of time on task scheduling and transmission. In addition, fine-grained tasks are allocated to high computing performance. The resource nodes will obviously reduce the resource utilization rate. In this cloud scenario, the dynamic resource scheduling strategy based on resource priority is implemented and the fine-grained tasks are integrated after grouping, and the bandwidth status of the resources is taken into consideration, and the scheduling strategy is optimized.
Secondly, the current scheduling strategy of cloud resource scheduling is not ideal in the load balancing of resource nodes, and the load imbalance between nodes is easy to occur. Most of the existing scheduling algorithms do not take into account the difference between the user groups and cause the VIP users not to get better service. In order to solve the bottleneck of the above cloud system, the Min-Min scheduling algorithm is used as the base. Based on the research and analysis, the complexity of the algorithm is low and easy to be realized, but its short board lies in the uneven resource load. Therefore, the classic Min-Min scheduling algorithm is first evolved to make the load evenly distributed and ensure the quality of VIP level service.
Finally, the research content of the paper is summarized and stated. The research results of this paper are stated, and the system learning of resource scheduling technology is used as the basis for the future research direction.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.01

【参考文献】