基于OpenStack的高资源利用率Docker调度模型

在现有OpenStack云平台与Docker容器技术的集成方案中，基于容器初始资源请求的调度模型由于未充分考虑容器运行时的实际资源使用情况，导致资源利用率较低。为满足云计算领域的高资源利用率和低成本需求，构建基于OpenStack云平台的Docker调度模型（DSM），将其与OpenStack的Keystone、Glance以及Neutron组件的API进行交互，获取创建容器所需的镜像、网络等资源，同时调用Docker Engine提供的API部署容器，对容器生命周期进行高效灵活管控。通过融合初始化模块、资源实时感知模块、容器调度模块、资源实时监测模块和容器迁移模块，并在容器调度模块中利用资源可用度评估与优先级决策调度机制为容器选择最优的计算节点，实现OpenStack云平台中资源的高效利用。实验结果表明，与经典Nova-Docker和Yun集成方案采用的调度模型相比，DSM调度模型在CPU和内存利用率上至少提升38.54、30.17个百分点和38.40、28.69个百分点。

OpenStack-based Docker Scheduling Model with High Resource Utilization

The existing integration schemes of the OpenStack cloud platform and Docker container technology adopt a scheduling model based on the initial resource request of the container, which does not fully reflect the actual resource usage of the container when running and results in low resource utilization.This study proposes a Docker Scheduling Model(DSM) based on OpenStack to satisfy the high resource utilization and low-cost requirements in cloud computing.The DSM interacts with the Application Programming Interfaces(APIs) of OpenStack's Keystone, Glance, and Neutron components to obtain resources, such as images and networks required to create containers.It deploys containers by calling the API provided by the Docker Engine to efficiently and flexibly manage the life cycle of containers.The DSM integrates the initialization, real-time resource awareness, container scheduling, real-time resource monitoring, and container migration modules.In addition, the DSM adopts Resource Availability-evaluation and Priority Decision-making(RAPD) scheduling mechanisms in the container scheduling module to select the optimal compute node for the container and efficiently utilize resources in OpenStack.The experimental results show that compared with the scheduling model used in Nova-Docker and Yun, the DSM improves CPU utilization by at least 38.54 and 30.17 percentage points, respectively, and improves memory utilization by at least 38.40 and 28.69 percentage points, respectively.