时空大数据环境下移动云计算的数据与服务迁移研究

Data and service migration is one of the most effective approaches to minimizing access latency, network workloads, and thus improving the quality of services for mobile cloud computing. However, due to the characteristics of mobile access patterns in big data environments in terms of diversity, real-time and large volumes, traditional data and service migration is facing severe challenges. By taking the advantages of mobile user's spatio-temporal trajectory data, this project intends to address four fundamental scientific problems in data and service migration with respect to its basic theory and implementation technologies: 1) completing the studies on the intractability of the data migration problem in its general forms to form the theoretical basis for the algorithm designs; 2) proposing more effective migration algorithms to extend the model by optimizing its supports to data storage and accesses; 3) proposing a concept of co-operative routing for service migration and investigating its impact on the network workloads and the total service costs, then introducing a traffic-aware migration algorithm; 4) studying the distribution of user requests whereby the merits of both online and off-line migration algorithms are combined to present an optimal algorithm to service the user accesses in an expected way. This work not only has great significance on the developments of mobile cloud applications, but also has great values for building highly cost-effective cloud infrastructures.

数据与服务迁移是减少延迟和网络负载，提高移动云计算服务质量的一种有效方法。然而面对大数据环境下用户访问云服务模式的多样性、实时性、巨量性等特点，目前的迁移策略和方法面临严重挑战。本项目基于大数据对移动访问的时空轨迹的挖掘信息，从基本理论和实践的两个角度研究数据与服务迁移的四个基础科学问题: 1) 完善一般形式下数据迁移问题难解性的证明，奠定算法设计的理论基础；2) 通过优化对数据存储和访问的支持，提出更加有效实用的迁移算法，拓展数据迁移模型的实用性；3) 针对服务迁移，提出合作路由的概念，研究其对网络负载和总服务成本的影响，藉此设计流量感知的迁移算法，使迁移更加有效；4) 研究用户访问依时间变化的统计分布，在此基础上，结合离线和在线迁移各自优势，提出以期望最优的方式满足以一定统计分布的用户访问的迁移算法。本项目对促进大数据时代移动云计算应用的发展和高性价比基础平台的构建具有重要的理论价值。

本项目拟以满足移动云计算中用户对服务的需求为目标，从基本理论和实现方法 两个角度针对现有研究中的四个未很好解决的问题分别展开紧密联系但又侧重点不同的四个方面的研究: 提高数据迁移模型的实用性、优化对数据存储、访问的支持、提高服务迁移决策计算的有效性、提高服务迁移算法的适应性。并在此基础上采用实验仿真和原型系统相结合的方法测试与确认了相关理论研究结果的正确性和通用性，通过不断的实验反馈提高和改进理论研究成果，优化原型系统及其关键性能指标。本项目针对上述四个问题开展的探索性研究工作在推动我国移动云计算向更深层次的迈进具有非常重要的基础研究价值。