【作者】 朱红宁；

【导师】 张斌；

【作者基本信息】 东北大学 ， 计算机应用技术， 2009， 博士

【摘要】 Web服务是独立的、模块化的、能够在网络上被描述、发布、发现和调用的应用程序。服务组合通过动态地合成已有的Web服务来实现更为强大的业务功能,从而为企业间的业务合成提供了无限的可能。近年来,网络中Web服务的数量不断扩大,如何从大量功能相同的Web服务中选出一组使组合流程有较好服务质量的服务,已经成为人们关注的热点。由于服务选取结果直接依赖于组合服务QoS模型的评价结果,因此,如何根据不同用户对不同业务流程中不同任务的个性化需求为组合流程建立合适的QoS模型,以使所选取出的服务具有较高的用户满意度就成为一个亟待解决的问题。非功能属性描述了Web服务在完成其功能的过程中表现出的执行特性,其取值情况直接影响它的服务质量(QoS)。人们一般用QoS指标表示对非功能属性拆分、合并、量化等处理后的指标,并使用服务的QoS来量化地表示其服务质量的优劣。因此,对于Web服务非功能属性的相关研究是QoS评价的基础。Web服务非功能属性具有主观性、业务相对性、相关性和不确定性,这些特性是本文研究的出发点。由于参与服务质量评价的QoS指标不止一个,通常的多属性综合评价理论是将各指标值加权,然后将多指标值合成为一个单值处理。从形式上看,QoS模型包含参评QoS指标以及各指标权重两部分内容。因此,本文重点探讨了QoS评价的不确定性处理、相关性分析与QoS指标、权重之间的密切关系。进一步,针对Web服务QoS评价中的基本问题,提出了支持服务选取的Web服务QoS评价方法。Web服务组合是支持业务流程逻辑的一组Web服务,其本身既可以是最终的应用,也可以是新的Web服务,组合通过确定不同Web服务的执行顺序和Web服务之间的复杂交互来实现。一般地,组合服务才‘能满足用户的需求。因此,本文提出了面向业务功能约束和相关性弱化的个性化QoS评价方法,它是一个针对基本服务的QoS评价模型,旨在消除一致性QoS评价的僵化。并在此基础上,提出了由基本服务的个性化QoS合成的而向组合业务的组合服务QoS评价模型。本文主要完成了如下的工作：(1)针对Web服务非功能属性不确定性,研究了不确定性的分类。进而,将Web服务评价指标的不确定性分为QoS指标的不确定性和指标权重的不确定性。其中,QoS指标的不确定性可分为描述的不确定性和行为的不确定性；指标权重的不确定性可分为主观不确定性和客观不确定性。提出了基于模糊集的QoS指标描述不确定性的处理方法；提出了基于灰预测的QoS指标行为不确定性的处理方法；提出了基于层次分析方法的主观权重不确定性的处理方法。(2)针对Web服务非功能属性之间的相关性,研究了相关性的分类,进而,将Web服务评价指标的相关性分为QoS指标之间的相关性和QoS指标与非功能属性约束之间的相关性。提出了基于灰关联分析的Web服务QoS指标之间相关性的分析方法；提出了基于联系数和集对势的Web服务QoS指标与非功能属性约束之间相关性的分析方法。(3)针对集对分析中只有联系权信息、没有特征权信息的特点,研究了特征加权集对分析,提出了支持服务选取的特征加权集对分析方法,使基于联系数和集对势的Web服务QoS指标与非功能属性约束之间相关性的分析方法变的更加细腻。(4)针对加权求和QoS评价方法中权重敏感问题,提出支持服务选取的QoS弃权评价方法。即直接比较Web服务QoS指标与非功能属性约束之间的关联程度,结果直接来自客观数据,排除了权重确定的困扰。(5)针对在组合流程的并行结构中,承担非关键任务的Web服务QoS被忽略的问题,研究了并行结构中关键路径的识别和非关键任务的获取方法,研究了依赖指标与相关指标之间的关系,提出了单目标满足、多目标优化的面向并行结构中非关键任务的QoS评价方法。(6)针对一致化QoS模型没有充分考虑Web服务非功能属性评价的主观性、业务的相对性、彼此的相关性和不确定性等特征,提出了面向业务功能约束和相关性弱化的个性化Web服务QoS评价模型,并基于此,提出了由基本服务的个性化QoS合成面向组合业务的组合服务QoS的模型。更多还原

【Abstract】 Web Services are independent, modelizational applied programs self-contained, modular applications that can be described, published, located, and invoked over a network. Services composition achieve more powerful business functions by combining existed web services dynamically and offer unprecedented opportunities for different business integrations in enterprises. As the number of web service increases, it’s getting hotter and hotter on how to select a group of web services among a mass of candidates with the same function so that the composite web service can achieve a fine quality of performance. Moreover, the service selection is dependent on the evaluation of composition services QoS model directly, so it becomes one of the urgent problems to construct a proper QoS model according to the individuation desire of different users for each task in the business flow, in order to select a group of more satisfied services.Non-functional properties describe the executive feature (e.g. execution time, reliability, availability and security etc) which will be shown in the process of completing the function by the Web Services, and the executive feature value influences the quality of service (QoS) directly. In general, people use the QoS cretiria to represent the ones of non-functional properties processed by separation, merge and quantization, and to indicate quantitatively good or not about the Web Services quality with the QoS. From above, we can see that the researches on Non-functional properties of Web services are the foundation of QoS evaluation. And this paper starts from such features as evaluation subjectivity, business relativality, dependency and uncertainty which are belong to Non-functional properties of Web services.There are many evaluation criteria on QoS, so multiattribute synthesis evaluation theories commonly offer each criterion a weight value, and then treat many criteria values as a single value. QoS model formally includes QoS cretiria and different cretiria weight values, therefore, we emphasis on uncertainty disposal of QoS evaluation, dependency analysis and the strong relationship between QoS cretiria ang their weights. Moreover, we provide a few of methods for QoS evaluation of Web services, which support services selection aiming at essential problems in QoS evaluation. Of Web servicesWeb services composition is a series of Web services which can both be the final application and be a new Web service. In practice, the execution sequence of different Web services and the complex interaction of Web services determine the composition. In general, only composite services can meet the users’requirement. Therefore, the paper supports individuation QoS evaluation method facing to the business functional constraints and relativity feebleness, and in fact, the method is a QoS evaluation model aiming at basic services in order to deliminite the rigidification of consistency of QoS evaluation, furthermore, and then builds up the composition services QoS evaluation model(orienting to composition services bussiness) integrated by individuation QoS compostion. We completed such works as following:(1) Aiming at uncertainty of non-functional properties of Web services, we research the uncertainty classification, and devide uncertainty of evaluation criteria of Web services with uncertainty of QoS criteria and uncertainty of criteria weights, therinto, and the former can be parted by uncertainty of description and behavior; and the latter can be separated by uncertainty of subjectivity and objectivity. And we provide a processing way for description uncertainty of QoS criteria based on fuzzy sets, a processing method for behavior uncertainty of QoS criteria based on grey forecasting, and a processing approach on subjectivity weight based on AHP.(2) Aiming at relativity of non-functional properties of Web services, we research the relativity classification, and devide relativity of evaluation criteria of Web services with relativity of QoS criteria themselves and relativity of QoS criteria and constraints of non-functional properties of Web services. Moreover, we support an analysis way for the relativity of QoS criteria of Web services based on grey relational analysis,and an analysis method for relativity of QoS criteria of Web services and constraints of non-functional properties based on connection-number and set-pair shi.(3) Aiming at features that there is not character weighted information, but only connection-number weight information in set pair analysis, we research the character weighted in set pair analysis, and bring up a character weighted set pair analysis way supporting services selection, which can turn the relativity of QoS criteria and constraints of non-functional properties of Web services into more acute.(4) Aiming at such sensitive problems as weighted sum and QoS evaluation method, we provide QoS non-weighted evaluation approach which supporing services selection, that is directly compare the relational degree between QoS criteria of Web services and constraints of non-functional properties of Web service, which can eliminate the confusion about weights and abtain the results from objective data straightly. (5) Aiming at such uncared-for problems as QoS of Web services working on non-critical tasks in composition flow, we research on a method for indetifying critical route and acquiring non-critical task in parallel structure, the relationship between dependency and relative criteria, and provide the QoS evaluation approach orienting to non-critical task in parallel structure with single goal satisfication and multi-goals optimizationy.(6) Aiming at such features as subjectivity of evaluation of non-functional properties, business relativity, dependency and uncertainty which are not considered completely in Web services, we provide that individuation QoS evaluation model of Web services orienting to business functions constraints and dependency weakening, and which is the foundation of the compositon services QoS model combined by individuation QoS of basic services of Web services orienting to compositon business.更多还原