【作者】 吴志强；

【导师】 吴长泽；

【作者基本信息】 重庆大学 ， 计算机应用技术， 2015， 硕士

【摘要】 Web服务的分布性、异构、自治和动态变化等特点,引发了服务计算环境的松散化、独立性、异构性等多种不确定因素,从而对服务提供者的非功能属性,即服务质量(Quality of Service,Qo S)提出更严格的要求。随着软件系统复杂度和人们对软件系统非功能属性要求的不断提高,保障Web服务的Qo S成为平台设计者和服务应用开发者所面临的重要问题。对于Web服务的服务质量保障,很多研究者采取建立Qo S模型对Web服务进行管理。已有的Qo S管理模型能够对Qo S数据进行基本管理,但是服务运行过程中无法实时更新动态的Qo S。这将导致Qo S数据失真,无法及时更新服务运行过程中的Qo S保障策略。为了解决Qo S模型无法有效保障Web服务运行过程中的动态Qo S,很多研究者提出Web服务动态评估方法来实时评估Web服务运行过程中的Qo S。通过实时搜集,计算Web服务运行过程各种属性数据,可以提供给用户及时、精确的Qo S数据信息,进而可以采取动态的服务质量保障措施来保障Web服务的服务质量。但是这种动态Qo S评估方式需要实时搜集服务器端,用户端数据,不但耗时,而且极大地增大了系统的开销。为了克服传统Web服务质量保障方法的不足,本文将态势感知技术引入Web服务质量保障领域,建立由状态和趋势组成的Qo S态势图,并根据当前态势采取合适的阶段性服务质量保障措施。不但完成了服务质量保障的动态性要求,而且降低了系统的开销。同时,通过提取当前态势数据中的灾变因子,预测Web服务未来的灾变情况,以便于提前采取相应适当的策略保障服务质量,进一步保障Web服务的服务质量。态势感知概念自从1985年提出后,被广泛应用于数据融合领域中,尤其是在网络安全感知领域。态势感知有多种模型,其中最常用的模型可以被概括为四个步骤(1)搜集感知领域原始数据;(2)通过相应的态势感知算法整理分析上一步收集的原始数据,产生感知领域的态势图;(3)结合上一步生成的态势图去进行更深层次的态势提取;(4)预测感知领域未来的发展态势。本文的研究重点就是将态势感知技术引入Web服务质量保障研究领域,进行Web服务Qo S态势感知以及预测。通过构建Web服务态势感知模型,完成Web服务当前Qo S态势的提取与识别,并根据当前态势给出合适的服务质量保障策略。然后通过对未来变点时刻的预测以及综合分析,提前采取积极主动的应变策略,从而更好地保障Web服务运行时的服务质量。本文的主要工作:①本文对Web服务领域基础知识和核心技术以及服务质量保障方法进行了深入的研究。针对目前服务质量保障领域存在的不足,本文将态势感知技术引入Web服务研究领域,建立Web服务态势感知模型。Web服务态势感知模型分为三个层次,分别是Qo S态势提取、Qo S态势识别以及态势预测。同时构建了由蕴含,等价,插件满足以及不满足四种状态,渐变点,突变点,半灾变点以及灾变点四类变点组成的态势感知指标体系。借助该指标体系衡量Web服务当前的状态和未来的发展趋势,进而采取积极主动的服务质量保障措施。②在构建的Web服务态势感知模型中,Qo S态势感知分为状态和趋势的感知。状态的感知可以通过与系统划定的属性基本满足区间进行比较得出。而趋势的识别需要通过最小二乘法以及决策树算法来求得。最终建立了由三种基元组成的Qo S态势图。在最小二乘法进行趋势提取过程中,当斜率p比较大的时候,累积和积累得相对较快,会更快到达阈值,降低趋势识别的效果。因此,提出一种根据斜率p分级设定阈值的解决方案来进行Qo S趋势识别。同时为进一步提升趋势识别的效果,本文给出一种分级遗传算法自适应阈值设定方法。通过经验设定趋势提取与识别算法中三个阈值取值范围,然后利用遗传算法的寻优特性,找到最小拟合误差的三个阈值。③通过趋势识别技术得到由三种基元组成的Web服务Qo S态势图。传统的Web服务Qo S预测方法要求大量的历史数据,且无法预测Qo S态势。而灰色灾变预测要求较少的数据,原理直观,计算复杂度合适,而预测效果不错。本文结合灰色预测这一显著特点,在得到Web服务状态集合和趋势集合基础上,给出渐变点、突变点、半灾变点、灾变点四类变点的定义。然后提出了基于GM(1,1)预测模型的变点预测方法。通过对未来一段时间内各类变点的预测及其组成的态势,通过采取阶段性Qo S保障策略,来保障Web服务运行时的Qo S。更多还原

【Abstract】 Distributed, heterogeneous, autonomous and dynamic change characteristics of Web services cause many uncertain factors of service computing environment such as loose, independence, and heterogeneous, which make stricter requirements of non functional properties of the service providers, namely, the service quality of service(Qo S). With the continuous improvement of complex degree and non functional properties requires of system software, guarantying the Qo S of Web services has become an important issue facing platform designers and application developers.For the service quality assurance of Web service, many researchers take the establishment of Qo S model to manage the Web service. The existing Qo S management model can manage the Qo S data, but the dynamic Qo S cannot be updated dynamically during the service operation. That will cause the Qo S data distortion, and can’t update the Qo S safeguard strategy in the service operation process. In order to solve the dynamic Qo S in the process of Qo S service operation, many researchers propose Web service dynamic assessment method to real-time evaluate the Web service running process Qo S. In order to solve the dynamic Qo S of Web service runtime, many researchers propose Web service dynamic assessment methods to evaluate the Web service runtime Qo S in real-time. Through real-time collection and computing various attribute data of Web service running process, users can get timely and accurate Qo S data information. And then it can take real-time service quality assurance measures to protect the service quality of Web service. However, this dynamic Qo S assessment method needs to collect the server side, the user data, which not only consumes time, but also greatly increases the overhead of the system.In order to overcome the disadvantage of Qo S guarantee method in Web service, situation awareness technology is introduced to Qo S security. Through the perception of the current status and future trends of Web services, replacement service and other measures can be taken before the decline or failure of Web services, so as to ensure the quality of the service. Since it was proposed in 1985, situation awareness has been widely used in data fusion field, especially in the field of network security awareness. The most commonly used model of situation awareness can be summarized into four steps :(1) Collect original data of perception field;(2) analysis the original data with the help of situation awareness algorithm, perceiving situation map;(3) Combining with the situation map generated by the previous step to extract a deeper trend;(4) to predict the future development trend perception. The focus of this paper is to introduce awareness technology to Qo S security research, completing Qo S situation awareness and forecast. Through the construction of situation awareness model, the extraction and recognition of current situation of service component is completed, and service quality assurance strategy are given according to the current situation.The main work of this paper:In this paper, the core technology of Web service and service quality assurance strategy are studied deeply. Aiming at the existing problems of the field of service quality guarantee, the situation awareness technology was introduces into the research field to establish situation awareness model of Web service. Situation awareness model is divided into three parts, respectively trend extraction recognition and forecast. At the same time, by constructing situation awareness index system which contain implication, equivalence, plug-in satisfaction and not meet the four conditions and four kinds of variable points, current status and future development trend of Web service were measured, and active measures were taken to guarantee quality of service.In the construction of situation awareness model, Qo S trend extraction is divided into state and trends. The extraction of state can be obtained by comparing with the Qo S basic meet interval designated by the system. While the trend extraction need to be solved by least squares method and decision tree algorithm. Finally, Qo S situation map containing three kinds of element is established to obtain the dynamic Qo S of component service. In the process of recognition by the least squares trend, when the slope of P is large, the accumulation increases quickly relatively, quickly reaches the threshold, decreased recognition effect. Therefore, a new trend identification solution was put forward according to the grade of slope P. At the same time, this paper presents a hierarchical genetic algorithm of adaptive threshold setting method to further enhance the accuracy of trend recognition. Set three threshold value range of trend recognition through experience, and then use the optimization characteristics of genetic algorithm to find three thresholds with the minimum fitting error.Qo S situation map can be obtained with the help of trend recognition. The traditional Web service Qo S prediction methods require a large amount of historical data, only for conventional Qo S forecast. The grey forecasting requires less data, intuitive principle, fit computational complexity and good prediction. Based on those traits of grey forecast, definition of four kinds of variable point was given on the basis of Qo S status and trend of component services. Through the forecast of the change point in the next period time and comprehensive analysis, the decisive measure of replacing the corresponding component services was taken in order to guarantee the quality of service components and achieve the purpose of service composition security.更多还原