基于LXC的PaaS云中支持QoS的自适应部署机制研究

发布时间：2018-06-02 09:17

本文选题：PaaS + LXC虚拟化　；参考：《青岛大学》2017年硕士论文

【摘要】：PaaS(Platform as a Service)云平台是一个由硬件基础设施与软件系统构成的、分布式的计算机集群系统。用户可以使用PaaS云平台上配置的资源开发和部署应用服务程序,并管理应用程序的执行。LXC(Linux Container)容器技术是操作系统级别的轻量级虚拟化技术,它为构建PaaS云平台带来了新契机。由于PaaS云平台是一个开放的、极其复杂的分布式运行环境,因此在平台上运行服务的执行环境与单机的执行环境有很大的不同。主要体现在对应用服务的分析、部署、监控等平台管理的运维方面。因此,如果还保持单机上部署应用服务一样的手工操作,不仅费时而且容易出错,因为在云平台上部署一个服务需要经过一长串复杂的配置操作,即便是有经验的开发者也会在修改大量配置文件时出现疏漏或重复,如果配置冲突将导致服务无法正常运行。尽管许多平台也简化了配置过程,但仍然需要服务开发者或PaaS平台提供者进行手工配置操作。针对上述问题,本文提出使用LXC容器构建PaaS云平台,以降低平台开销,提高平台的整体性能;并在该PaaS云平台上设计了一种支持QoS的自适应部署机制模型,该模型根据云平台提供商和用户之间签署的服务等级协议SLA为用户选择满足其服务质量QoS要求的部署节点,同时基于负载均衡策略进行应用服务的部署。具体工作如下:首先,在分析研究LXC的Namespaces和Cgroups技术的基础上,提出一种采用LXC虚拟化技术构建一个简易的轻量级PaaS云平台的方法,以达到隔离不同租户和共享云平台软硬件资源的目的;并用相关实验证明该方法相比传统虚拟机方法更具性能优势,更适合于提供科学计算服务的PaaS云平台。其次,设计PaaS云平台服务部署节点选择优化算法,以实现对应用服务的部署和运行。在分析影响平台节点选择的QoS参数和随机负载均衡策略的基础上建立目标函数。以节点的当前负载阈值和服务部署请求的QoS属性值作为约束条件,使用混合整数线性规划建模并求解。该算法可自动实现应用服务的部署任务。最后,在服务器集群上构建了基于LXC的PaaS云平台,并在该平台上设计实现了支持QoS的自适应部署机制模型。并通过系统测试验证平台及所做研究工作的有效性和可行性。最后对本文工作进行总结和展望。
[Abstract]:PaaS(Platform as a Service) cloud platform is a distributed computer cluster system composed of hardware infrastructure and software system. User can use the resources configured on PaaS cloud platform to develop and deploy application service program, and manage the execution of application program. LXCU Linux container technology is a lightweight virtualization technology at operating system level, which brings a new opportunity to build PaaS cloud platform. Because the PaaS cloud platform is an open and extremely complex distributed running environment, the execution environment of running services on the platform is very different from that of the single machine. Mainly embodies in the application service analysis, the deployment, the monitoring and so on platform management operation and maintenance aspect. Therefore, if you keep the same manual operation as an application service deployed on a single machine, it is not only time-consuming but also error-prone because deploying a service on a cloud platform requires a long and complex set of configuration operations. Even experienced developers may omit or duplicate changes to a large number of configuration files, which will cause the service to fail in case of configuration conflicts. Although many platforms also simplify the configuration process, manual configuration is still required by service developers or PaaS platform providers. To solve the above problems, this paper proposes to use LXC container to build PaaS cloud platform to reduce the overhead of the platform and improve the overall performance of the platform, and design an adaptive deployment mechanism model supporting QoS on the PaaS cloud platform. According to the service level protocol (SLA) signed between the cloud platform provider and the user, the model selects the deployment nodes that meet the requirements of the quality of service (QoS), and deploys the application service based on the load balancing strategy. The main work is as follows: firstly, on the basis of analyzing the Namespaces and Cgroups technology of LXC, a simple method of constructing a lightweight PaaS cloud platform using LXC virtualization technology is proposed. In order to isolate different tenants and share the hardware and software resources of cloud platform, the experiments show that this method has more performance advantages than traditional virtual machine method, and is more suitable for PaaS cloud platform which provides scientific computing services. Secondly, the PaaS cloud platform service deployment node selection optimization algorithm is designed to implement the deployment and operation of application services. The objective function is established based on the analysis of the QoS parameters and random load balancing strategies that affect the platform node selection. Using the current load threshold of the node and the QoS attribute value of the service deployment request as the constraint conditions, the mixed integer linear programming is used to model and solve the problem. The algorithm can automatically implement the deployment of application services. Finally, the PaaS cloud platform based on LXC is constructed on the server cluster, and an adaptive deployment mechanism model supporting QoS is designed and implemented on the platform. The validity and feasibility of the platform and the research work are verified by the system test. Finally, the work of this paper is summarized and prospected.
【学位授予单位】：青岛大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP393.09

【参考文献】