面向可靠性的微服务系统自适应调整技术研究

发布时间：2018-12-10 07:03

【摘要】：微服务架构是当前最流行的软件架构,当开发者从微服务架构获得敏捷开发时,可靠性成为系统运行的最大的痛点。本文重点研究了面向可靠性的微服务系统自适应调整中的两个重要问题:微服务系统可靠性分析和服务流程组合。本文在对软件系统可靠性进行研究的基础上,提出了一种软件系统可靠性分析任务划分的框架;基于此框架介绍了四类常用的可靠性分析技术以及相应的方法,调查了将这些技术和方法用于分析软件系统可靠性的研究进展,发现了每种可靠性分析任务与相应的技术、方法之间的关联;在此基础上,重点关注了基于概率模型检验的可靠性分析方法以及这类分析方法的应用场景,讨论了概率模型检验方法在分析软件系统可靠性时所具有的优势,并给出了一个通用的基于概率模型检验的软件可靠性分析平台。微服务系统是与单体架构相对应的一个概念。传统的单体架构多实例部署机制在应对大流量、高并发用户洪水般访问时,有效缓解系统压力的同时,也造成了极大的资源浪费。微服务架构以其既能缓解系统压力又能充分利用系统资源的优势成为系统架构的首选。传统软件系统的可靠性分析方法都是针对单体多实例部署的情况,使用传统的可靠性分析方法不能准确地分析微服务差异化多实例部署模式下的微服务系统的可靠性。针对这一问题,本文提出了一种微服务差异化多实例部署模式下微服务系统可靠性分析方法,实验验证了这种方法的有效性。随着微服务系统业务规模的扩大,单一的服务流程已经不能满足微服务系统业务的扩张,服务流程的多元化已成微服务系统发展的必然趋势。然而,用户在使用微服务系统时,很少完全使用微服务系统提供的全部业务,而仅使用其感兴趣的部分业务,在这种情况下,使用传统的单一业务流程的可靠性分析方法不能准确地反映多业务流程的微服务系统的可靠性,同时,多元化业务流程的业务交织增加了微服务系统可靠性分析的难度。针对这一问题,本文提出了一种多社区微服务系统可靠性分析方法,实验验证了这种方法的有效性。服务流程重组是实现微服务系统自适应调整的一种最直接有效的方法。传统的服务流程组合问题的解决方案都是以服务与服务模型粒度相同为前提,事实上,在互联网中存在这样一类大粒度服务,即,一个服务可以完成多个服务模型描述的多个需求规约,与使用和服务模型粒度相同的服务组合可靠性最优的服务流程相比,使用大粒度服务组合服务流程通过减少参与组合服务流程的候选服务集合的数目,提高组合服务流程的效率。本文通过解决对最优大粒度服务流程建模和组合可靠性最优的大粒度服务流程两个子问题完成大粒度服务流程组合问题问题,并通过实验验证了这种方法的有效性。最后,参考IBM提出的自治元素模型,本文提出了一种面向可靠性的微服务系统自适应调整框架,并利用本文的研究成果及分布式集群技术和大数据技术实现了该框架的原型平台。在云计算环境中,该自适应调整平台利用集群技术及大数据技术以及基于对多实例部署、多业务流程的微服务系统的可靠性分析及面向大粒度的服务流程组合方法,实现面向可靠性的微服务系统自适应调整。
[Abstract]:The micro-service architecture is the most popular software architecture, and when the developer obtains the agile development from the micro-service architecture, the reliability becomes the biggest pain point of the system operation. This paper focuses on two important problems in the self-adaptive adjustment of the reliability-oriented micro-service system: the reliability analysis of the micro-service system and the combination of the service flow. Based on the research of the reliability of the software system, this paper presents a framework of the reliability analysis task division of the software system. Based on this framework, four kinds of common reliability analysis techniques and corresponding methods are introduced. The research progress of these techniques and methods to analyze the reliability of the software system is investigated, and the correlation between each reliability analysis task and the corresponding technology and method is found. On this basis, The reliability analysis method based on the probability model test and the application scene of this kind of analysis method are focused on, and the advantages of the probability model test method in analyzing the reliability of the software system are discussed. A general software reliability analysis platform based on probability model test is given. The micro-service system is a concept corresponding to the monomer architecture. The traditional multi-instance deployment mechanism of the single-body structure can effectively relieve the pressure of the system when the large-flow and high-concurrency user is in flood-like access, and also causes great resource waste. The micro-service architecture can only relieve the system pressure but also make full use of the advantage of system resources as the first choice for the system architecture. The reliability analysis method of the traditional software system is based on the multi-instance deployment of the monomer, and the reliability of the micro-service system under the multi-instance deployment mode of the micro-service can not be accurately analyzed by using the traditional reliability analysis method. In order to solve this problem, this paper presents a method for analyzing the reliability of the micro-service system under the multi-instance deployment model of the micro-service differentiation, and the effectiveness of this method is verified by the experiment. With the expansion of the business scale of the micro-service system, the single service flow cannot meet the expansion of the service of the micro-service system, and the diversification of the service flow has become the inevitable trend of the development of the micro-service system. However, when a user is using a micro-service system, the entire service provided by the micro-service system is rarely fully used, and only part of the service of which it is of interest is used, in which case, The reliability analysis method of the traditional single business process can not accurately reflect the reliability of the micro-service system of the multi-service process, and meanwhile, the service interleaving of the diversified business process increases the difficulty of the reliability analysis of the micro-service system. In this paper, a multi-community micro-service system reliability analysis method is proposed, and the effectiveness of this method is verified. Service process reengineering is one of the most direct and effective ways to realize the self-adaptive adjustment of the micro-service system. the solution of the traditional service flow combination problem is based on the same granularity of the service and the service model, in fact, there is a large size service in the internet, that is, a service can complete a plurality of demand protocols described by a plurality of service models, The efficiency of the combined service flow is improved by reducing the number of candidate service sets participating in the combined service flow by reducing the number of candidate service sets participating in the combined service flow, as compared to a service flow that uses the same service combination reliability as the service model granularity. In this paper, the problem of large-size service flow combination is solved by solving the two sub-problems of the optimal large-size service flow modeling and the optimal combination reliability, and the effectiveness of this method is verified by the experiment. Finally, with reference to the autonomous element model proposed by IBM, this paper presents a self-adaptive adjustment framework for the reliability-oriented micro-service system, and the prototype platform of the framework is realized by using the research results of this paper and the distributed cluster technology and the big data technology. in a cloud computing environment, the adaptive adjustment platform utilizes the clustering technology and the large data technology and the service flow combination method for the large-size-oriented service flow based on the reliability analysis of the micro-service system with the multi-instance deployment and the multi-business process, and the self-adaptive adjustment of the reliability-oriented micro-service system is realized.
【学位授予单位】：上海大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TP311.53

【相似文献】