移动云计算中基于资源感知的动态负载迁移策略研究

Offloading technology in mobile cloud computing is aimed at offloading tasks with heavy load from mobile terminals such as smartphones to cloud platforms. Thus, offloading can break through performance bottleneck brought by resource limitations of mobile terminals. The cloud platforms and mobile networks usually have dynamic computation resources and network resources. However, existing work does not consider the dynamic characteristics of resources overall in the course of offloading, and they have not studied the decline of QoS caused by the reduction of resources. In view of this, this project intends to make research on dynamic offloading strategy based on resource awareness. Firstly, we propose vehicular cloud integrated heterogeneous cloud service model to increase available resources and study the method of resource discovery, maintenance and usage of vehicular cloud. Secondly, we propose effectiveness optimization based service selection strategy. The QoS of cloud platforms is estimated through modeling, solving and optimizing the problem of offloading, and then service selection is done according to the estimated QoS. Lastly, we propose resource awareness based service handoff strategy. The invalid services are discovered through dynamic estimation on QoS, and service handoff will be performed once a service is invalid. This research is expected to improve the QoS of mobile cloud computing system further and provide technical support for breaking through performance bottleneck of mobile terminals.

移动云计算中的负载迁移技术旨在将移动终端（如智能手机）负载较重的任务迁移到云平台执行，从而突破移动终端资源局限性所带来的性能瓶颈。云平台和移动网络通常具有动态的计算资源和网络资源，然而已有的工作没有全面考虑在负载迁移过程中资源的动态特性，并且没有研究资源减少导致的服务质量下降问题。鉴于此，本项目拟研究移动云计算中基于资源感知的动态负载迁移策略。首先提出整合车载云的异构云服务模型以增加可用资源，并研究车载云的资源发现、维护和使用方法；然后提出基于效能优化的服务选择策略，通过对负载迁移问题进行建模、求解和优化来预估云平台的服务质量并进行选择；最后提出基于资源感知的服务切换策略，通过对云平台的动态服务质量预估来发现失效服务，针对失效服务进行服务切换。该研究可望进一步提升移动云计算系统的服务质量，并且为突破移动终端的性能瓶颈提供技术支撑。

本项目对移动云计算中基于资源感知的动态负载迁移策略展开了研究，研究成果主要包括以下几个方面：（1）提出了一种车载云环境下面向多用户的任务调度策略，该策略通过联合优化任务传输时间和执行时间来提升车载云的服务质量。（2）提出了一种基于效能优化的负载迁移策略，该策略能够满足应用截止时间，同时降低移动终端的能耗。（3）提出了一种基于资源感知的动态任务调度策略，该策略具有较高的服务成功率。（4）提出了一种基于强化学习的服务迁移策略，该策略在应用实时性和服务成本之间获得了较好的性能平衡。本项目的研究推动了移动云计算系统负载迁移技术的发展，具有重要的理论意义和应用价值。