云平台下多用户应用并发部署机制的研究与实现

发布时间：2018-09-18 15:53

【摘要】：未来网络研究一直以来都是计算机网络方面的研究热点。未来网络将为用户提供一个更加先进的网络环境,能够更好地支撑上层服务以及应用。然而目前很多科研人员在进行应用或服务的研究和实验时,由于自身资源或工具限制,难以搭建出面向未来网络的服务实验环境,只能在本地搭建低性能的网络环境。而且科研人员在搭建本地的环境的时候,需要手动部署各种软件,耗时耗力,严重影响了科研效率。针对这一现状,需要为用户提供一个能将资源、工具进行集中管理、部署的服务器创新环境,需要搭建一个基于自动化部署机制的服务创新云平台。但是随着服务创新云平台的使用用户的增多,部署机制在运行中将会面临着多并发对系统性能的挑战。用户的多并发请求将会导致响应滞后,系统性能下降甚至瘫痪崩溃。针对这些问题,本论文研究了面向多用户并发请求下的部署机制的系统架构的设计,从单台扩展、动态负载均衡扩展、分布式扩展等方面提出了优化部署机制的架构设计的方案,并结合OpenStack云平台和自动化部署工具Puppet提出三种架构下的部署机制,实现对应用的自动化部署、对已有应用的重新部署以实现应用版本自动切换等功能。本论文的主要研究内容如下:1.设计并实现了云平台下多用户并发部署的单台机制。包括使用Apache结合Passenger的形式来改变部署架构,并实现单机机制下的实验环境首次搭建的应用一键部署功能和需求变更下的应用重新部署功能。2.设计并实现云平台下多用户并发部署的动态负载机制。包括使用Nginx结合Puppet集群来改变部署架构,采用监控服务器性能来动态更改负载均衡策略,并实现动态负载均衡机制下的实验环境首次搭建的应用一键部署功能和需求变更下的应用重新部署功能。3.设计并实现云平台下多用户并发部署的分布式机制。包括使用Git结合Puppet来改变部署架构,实现服务器代码的自动推送、客户端代码的自动拉取,并实现分布式机制下的实验环境首次搭建的应用一键部署功能和需求变更下的应用重新部署功能。为验证多并发下每个机制下系统可用性,对云平台下的每个部署机制进行了功能测试和性能测试。功能测试表明,本文提出的基于云平台的三种部署机制都能实现预期的功能需求。性能测试表明,当并发请求较小时,三种部署机制都能提供优异的性能。但随着并发请求不断增加,单台机制的性能极速下降,动态负载机制的性能稍微有所下降,分布式机制则基本维持一个优异的性能。
[Abstract]:The future network research has always been the research hotspot of computer network. The future network will provide users with a more advanced network environment, can better support the upper layer of services and applications. However, in the research and experiment of application or service, it is difficult for many researchers to build a service experimental environment for the future network because of their own resource or tool constraints, and only build a low-performance network environment in the local area. When setting up the local environment, the researchers need to deploy all kinds of software manually, which is time-consuming and labor-consuming, which seriously affects the efficiency of scientific research. In view of this situation, we need to provide a server innovation environment which can centralize the management and deployment of resources and tools for users, and build a service innovation cloud platform based on automated deployment mechanism. However, with the increasing number of users using the service innovation cloud platform, the deployment mechanism will face the challenge of multi-concurrency to the system performance. Multiple concurrent requests from users can lead to latency, system performance degradation or even paralysis. Aiming at these problems, this paper studies the design of system architecture for deployment mechanism under multi-user concurrent request, from single platform to dynamic load balance expansion. In the aspect of distributed extension, the architecture design of optimizing deployment mechanism is put forward. Combining with OpenStack cloud platform and automatic deployment tool Puppet, three kinds of deployment mechanisms are put forward to realize the automatic deployment of application. The redeployment of existing applications to achieve automatic application version switching and other functions. The main contents of this thesis are as follows: 1. This paper designs and implements a single mechanism of multi-user concurrent deployment on cloud platform. Including the use of Apache combined with the form of Passenger to change the deployment architecture, and the implementation of the experimental environment under the single-machine mechanism, the first time to build the application one-click deployment function and the requirements of the application redeploy function. 2. The dynamic load mechanism of multi-user concurrent deployment on cloud platform is designed and implemented. Including the use of Nginx combined with Puppet cluster to change the deployment architecture, the use of monitoring server performance to dynamically change load balancing strategy, And implement the first application one-click deployment function and the application redeploy function .3under the dynamic load balancing mechanism under the experimental environment for the first time. The distributed mechanism of multi-user concurrent deployment on cloud platform is designed and implemented. Including the use of Git combined with Puppet to change the deployment architecture, the implementation of server code automatic push, client code automatically pull, The first application one-click deployment function and the application redeployment function under changing requirements are implemented in the distributed experimental environment. In order to verify the system availability under each mechanism with multiple concurrency, each deployment mechanism under the cloud platform is tested for function and performance. The functional tests show that the three deployment mechanisms proposed in this paper can meet the expected functional requirements. Performance tests show that all three deployment mechanisms provide excellent performance when concurrent requests are small. However, with the increasing of concurrent requests, the performance of the single mechanism is extremely fast down, the performance of the dynamic load mechanism is slightly decreased, and the distributed mechanism basically maintains an excellent performance.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP393.09

【相似文献】