基于云计算的资源受限环境应用加速框架研究

发布时间：2018-03-09 17:23

本文选题：Python　切入点：云计算　出处：《南京理工大学》2013年硕士论文　论文类型：学位论文

【摘要】：随着通信技术与计算技术的发展,计算与通信能力兼备的智能设备已经非常普遍,但是由于受到尺寸、价格以及能源的限制,这些智能设备的硬件资源(CPU,内存等)是相当有限,因此,智能设备在处理复杂应用的时候会有些力不从心。目前,云计算是的研究热点,它应用了成熟的虚拟化技术将大量网络连接的计算资源统一管理和调度,构成服务供互联网上的用户使用。利用云计算提供的资源可以为资源受限的智能设备提供本地应用加速的功能,从而提高用户体验,扩展设备的应用场景。目前现有的云计算平台大多针对PC平台设计,并且编程框架复杂,无法实现资源受限环境下应用的动态迁移。本文针对资源受限环境下的应用,设计了一种跨平台的、可以动态迁移应用的、编程模型简单的应用加速框架。本框架能够把异构环境下的应用程序中复杂任务自动地、无缝地迁移到云端,在云端自动部署执行,从而通过利用云端丰富的计算资源加速复杂任务的执行,来起到加速整个应用的执行的作用。本框架的设计采用了序列化、虚拟化与沙箱、分布式计算等技术,提供了细粒度的计算迁移策略、简易的用户接口,以及良好的可扩展性与容错性。本文最后采用压缩感知中的正交匹配追踪算法(Orthogonal Matching Pursuit, OMP)对框架进行了应用加速的测试,实践证明,本文设计的框架能对资源受限环境下的应用起到加速作用。
[Abstract]:With the development of communication technology and computing technology, intelligent devices with both computing and communication capabilities have become very common. However, due to the limitations of size, price and energy, the hardware resources of these intelligent devices are very limited, such as CPU, memory, etc. At present, cloud computing is a research hotspot. It uses mature virtualization technology to manage and schedule a large number of computing resources of network connection. Making use of the resources provided by cloud computing can provide the local application acceleration function for the intelligent device with limited resources, which can improve the user experience and expand the application scenario of the device. At present, most of the existing cloud computing platforms are designed for PC platforms, and the programming framework is too complex to realize the dynamic migration of applications in resource-constrained environment. In this paper, a cross-platform is designed for applications in resource-constrained environments. This framework can automatically and seamlessly migrate complex tasks in heterogeneous applications to the cloud, and automatically deploy and execute in the cloud. By using rich computing resources in the cloud to accelerate the execution of complex tasks, this framework is designed using serialization, virtualization and sandbox, distributed computing, etc. Fine-grained computing migration strategy, simple user interface, and good scalability and fault tolerance are provided. In the end of this paper, orthogonal Matching pursuit (omp) algorithm is used to accelerate the application of the framework. It is proved that the proposed framework can accelerate the application in resource-constrained environment.
【学位授予单位】：南京理工大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP3

【参考文献】