云资源弹性调度测试方法的研究与实现

发布时间：2018-06-10 17:52

本文选题：OpenStack + 弹性规则集　；参考：《内蒙古大学》2017年硕士论文

【摘要】：随着云服务的普及,越来越多的企业将软件系统迁移到云上,利用云平台的弹性服务来自动应对系统工作负载的变化,可以提高资源的利用率。弹性就是云平台随着系统应用工作负载变化来自动扩展和回收底层云资源。一个合适的云端资源购买方案对上层云应用而言是至关重要的,它包括选择怎样的弹性水平(自动扩展回收资源的策略)和购买多少云端资源。虽然大多数云服务提供商都说弹性是他们云平台拥有的优点,但对于云资源消费者来说,云端资源的具体使用情况不可知,无法考量所使用的云平台的弹性水平是否与系统负载很好匹配。因此我们期望能找到一种可行的测试方法,帮助用户测试出应用程序在弹性云平台中运行时最匹配的弹性水平和具体的资源使用数量。对于云资源消费者来说,最匹配的弹性水平代表最低的资源使用量,该测试结果可以提供最佳的云端资源购买方案,帮助用户使用最低的成本达到需求的系统性能;对于云服务提供商来说,最合适的弹性水平代表最好的弹性服务,该测试结果可以提供最佳的资源分配方案,使云端资源得到充分利用。本文从测试的角度提出并实现了一种可行的云资源弹性调度测试方法,主要工作包括:(1)搭建集成OpenStack与Docker的弹性云平台:OpenStack和Docker之间是很好的互补关系。两者的结合可以实现弹性规则集的自由设置,模拟多种弹性水平供应用测试。(2)基于CloudSuite标准测试应用程序套件的测试场景设计:根据用户的应用行为分为三种应用测试场景:基准应用测试场景、横向应用拓展测试场景、纵向应用拓展测试场景。根据各测试场景的要求编写HOT模版代码将应用编排部署到平台上,以此来模拟用户的云应用及其工作负载。(3)基于弹性规则集设置的测试例生成:规则集内容围绕监控间隔、上达阈值、下达阈值、上达持续时间、下达持续时间、冷却时间等参数组合设计,通过等价划分与边界值方法设计不同的弹性规则集,然后根据不同的应用测试场景,综合考虑规则集和虚拟机系统性能来设计测试用例。(4)弹性调度测试的执行和结果分析:利用OpenStack中的Ceilometer模块监控测试执行并收集测试数据写入日志,分析各测试例的资源调度情况,得出各种应用测试场景下弹性云端资源的最佳购买方案。
[Abstract]:With the popularity of cloud services, more and more enterprises migrate software systems to the cloud, and use the flexible services of cloud platform to automatically cope with the changes of system workload, which can improve the utilization of resources. Elasticity is a cloud platform that automatically expands and reclaims underlying cloud resources as the system application workload changes. A suitable cloud resource purchase scheme is very important for the upper cloud application. It includes the selection of the level of elasticity (the strategy of automatic expansion and recovery of resources) and the number of cloud resources purchased. While most cloud service providers say flexibility is an advantage of their cloud platform, for consumers of cloud resources, the exact use of cloud resources is unknown. It is not possible to consider whether the level of elasticity of the cloud platform used matches well with the system load. Therefore, we hope to find a feasible test method to help users test the level of elasticity and the specific amount of resources used when the application is running in the elastic cloud platform. For cloud resource consumers, the best matching elasticity level represents the lowest resource usage. The test results can provide the best cloud resource purchase scheme and help users to achieve the required system performance with the lowest cost; For cloud service providers, the most appropriate level of flexibility represents the best flexible service. The test results can provide the best resource allocation scheme and make full use of cloud resources. From the point of view of testing, this paper proposes and implements a feasible test method for elastic scheduling of cloud resources. The main work includes: 1) to build a good complementary relationship between OpenStack and Docker's elastic cloud platform: OpenStack and Docker. The combination of the two can achieve the free setting of the elastic rule set, Test scenario design based on CloudSuite standard test application suite: according to the user's application behavior, there are three application test scenarios: benchmark application test scenario, horizontal application extension test scenario, Vertical application extension test scenario. According to the requirements of each test scenario, write the hot template code to deploy the application choreography to the platform to simulate the user's cloud application and its workload. 3) the test case generation based on the flexible rule set setting: the rule set content revolves around the monitoring interval, Based on the combination design of parameters such as the threshold value, the duration, the duration and the cooling time, different elastic rule sets are designed by the method of equivalent partition and boundary value, and then according to the different application test scenarios, the method of equivalent partition and boundary value is used to design different elastic rule sets. Considering the performance of rule set and virtual machine system synthetically to design test case. 4) the execution and result analysis of elastic scheduling test: the Ceilometer module in OpenStack is used to monitor test execution and collect test data to write to the log. By analyzing the resource scheduling of each test case, the optimal purchasing scheme of elastic cloud resources under various application test scenarios is obtained.
【学位授予单位】：内蒙古大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP311.53;TP393.09

【相似文献】