面向云数据中心应用感知的参与式资源调度技术研究

Data centers are the base of service provisioning on cloud platforms. Resource scheduling in data centers should make efficient use of physical resources and guarantee the performance of applications deployed atop data center networks. Existing scheduling frameworks have already tried to support heterogeneous dynamic resource demands from cloud users, but did not effectively solve the challenge posed by dynamic resource demands from cloud users on resource scheduling in data centers. In this project, for virtualization-based cloud data centers, we aim at systematically developing an application-aware dynamic resource scheduling mechanism, and start from incorporating cloud users’ participation into the scheduling on the cloud platform side. More specifically, in our project, cloud users can send queries for running details of data centers and their applications based on well-defined models and interfaces; depending on the running statuses of applications, cloud users can also provide potential future changes of their resource demands for cloud platforms, which will be beneficial for cloud platforms when they make resource scheduling decisions (e.g., allocation, reservation, and migration); on the other hand, cloud platforms can change the charge policies on physical resources to encourage or force cloud users to adjust their future resource demand curves. We will also evaluate the aforementioned research outcome on the cloud testbed that was independently developed by the project team.

数据中心是云平台对外提供各类云服务的物理承载者。数据中心资源调度应高效利用底层物理资源、保证所部署的应用的稳定性能。已有的技术框架在适应异构动态的用户资源需求方面提供了支持，但没有较好地解决用户的动态资源需求对云平台资源调度的挑战。本课题拟面向基于虚拟化技术的云数据中心，以系统化构建应用感知的动态资源调度机制为目标，以云平台引入用户参与调度为切入点，实现云平台与用户协同的参与式资源调度。具体而言，基于良好定义的模型和接口，用户可主动询问数据中心物理机及网络的运行细节；根据应用的运行状态，用户可主动向云平台提供其资源需求可能的变更信息，便于云平台应用感知的资源分配、预留和迁移；云平台亦可基于激励机制向用户反馈物理资源的动态定价信息，鼓励或迫使用户调整资源需求曲线。基于课题组自主研发的云计算实验床，系统化测试与验证上述研究成果。

数据中心资源调度应高效利用底层物理资源、保证所部署的应用的稳定性能。本课题面向基于虚拟化技术的云数据中心，以系统化构建应用感知的动态资源调度机制为目标。基于该项目，项目团队在移动云环境下高效资源调度、协作能量调度、混合数据中心内虚拟网络部署、数据分析系统中基于半先验知识的可证明有效的调度、兼顾公平与效率的数据中心流调度、无线能量传输中充电器放置和能量分配的联合优化、异构边缘云资源调度、云-边缘系统中跨域大数据作业调度、协同交互充电等方面进行了深入的研究，取得了重要进展。发表或含录用论文21篇，专著1章；申请专利2项。获得2016年度江苏省优秀博士学位论文奖、2016年度南京大学优秀博士学位论文奖，江苏省科学技术奖一等奖（第9完成人）、还获得自然科学基金面上项目、江苏省自然科学基金面上项目、CCF-腾讯犀牛鸟科研基金、创意基金资助各1项。