面向计算密集型移动应用的朵云高效服务提供及无缝切换技术研究

随着基于移动设备的计算密集型复杂应用不断增长，研究如何在移动云中高效地为移动端用户提供优质服务具有重要的理论和应用价值。本项目在一种移动云计算模式——朵云环境下，针对移动端资源贫乏、移动应用类型多样及朵云服务范围有限等特点，研究面向计算密集型移动应用的朵云高效服务提供及无缝切换技术，以提升朵云服务质量、移动用户满意度和系统性能。确定弹性划分粒度、任务耦合度和均衡度，研究基于图聚类的上下文感知朵云应用弹性划分方法，提高应用划分精准度。建立机会卸载模型，定义卸载成本函数，研究基于广义Nash均衡的多朵云用户机会卸载方法，降低卸载成本及终端能耗。提出基于影子页表和副本限制的朵云实时任务双容错调度策略，提高朵云系统的可靠性。建议基于切换增益函数和TOPSIS的水平/垂直朵云服务切换方法，保证服务提供的连续性。构建基于校园移动网络的朵云实验环境，对所提出的方法进行测试，并提供校际朵云示范应用环境。

With the continuous growth of mobile devices based complex compute-intensive applications, the research on how to provide quality service for mobile users has very important theoretical significance and practical value. By analyzing features of mobile device resource insufficiency, mobile application diversity and cloudlet service scope limitation, this project studies the efficient cloudlet service provisioning and seamless service handoff for compute-intensive mobile applications to improve cloudlet's QoS, mobile user satisfaction and system performance. In this project, the graph clustering based context-aware cloudlet application elastic partitioning approach is studied by determining the elastic partitioning granularity, coupling degree of the task and the task balancing degree, which enables to improve the accuracy of application partitioning. For reducing the offloading cost and mobile terminal energy consumption, the multi-cloudlet user opportunistic offloading method based on generalized Nash equilibrium is studied by establishing opportunistic offloading model and defining offloading cost function. Moreover, the dual fault-tolerant scheduling for cloudlet real time tasks based on shadow page tables and replica limitation is proposed to enhance cloudlet system reliability. The horizontal/vertical cloudlet service handoff approach based on handoff gain function and TOPSIS method is proposed to guarantee the continuity of the service provided. The cloudlet environment based on campus mobile network is constructed, on which the simulation is performed for testing the presented methods. The interuniversity cloudlet demonstration environment is also provided.