Web服务组合关键机制

发布时间：2019-05-10 21:28

【摘要】：互联网上正具有越来越多的可以被获取和访问的网络资源,包括存储资源、计算资源、应用资源和软件资源,使得互联网逐渐成为一个开放的分布式计算平台。而面向服务的计算(SOC)和面向服务的架构(SOA)是当前分布式计算和软件开发的最新发展方向,能够帮助企业或互联网用户更加灵活、快速、低成本地开发应用程序和业务流程,其中,Web服务组合则是面向服务计算的关键技术。随着云计算、电子商务以及Internet等技术的发展,Web服务组合发生的环境发生了变化。为了实现普适性,很多基于QoS的Web服务组合方法针对通用环境而设计,没有考虑特定环境中面临的一些新问题与机会。本文将Web服务组合发生的环境进行了细分,对基于QoS的Web服务组合进行了研究,重点在于改进传统的方法使之适应这些特定环境、提出新的方法使之充分利用这些特定环境的优势并进一步研究特定环境带来的新问题,提出相应的解决方法。本文主要内容及贡献如下： 1.由于互联网是由具有不同延迟的子网构成,因此位于不同网络位置的用户对相同的Web服务在服务质量上的体验具有差异。然而,服务提供商一般只提供服务QoS的平均值作为服务的评价指标,造成了目前大多数服务组合方法没有考虑网络传输性能给不同位置的用户调用Web服务带来的影响。此外,在考虑到Web服务处理能力有限的情况下,同时使用多个实体服务满足多个用户的服务请求,不但能够提高组合服务的可靠性和抗毁容灾的能力,还能够显著降低Web组合服务的总处理时间。但是,已有的方法在这样的情况下对服务动态组合的研究不足。本文借助排队网络模型,引入了网络QoS的量化指标,给出优化变量的约束条件和上下限,提出了一种运行时服务组合方法。该方法首先使用非线性优化理论选择一组优势实体服务,再根据网络当前的实际状况,在运行时选择合适的候选服务。实验表明,本文提出的方法兼顾了解的最优性和执行效率两个方面,适合参数动态变化的情境。 2.以前的研究方法侧重对原子Web服务或集群服务在能源效率上的研究,却忽视了对整个组合服务的能耗研究。本文根据Web候选服务的能耗模型,计算整个组合服务的能耗,提出了能耗感知的服务组合方法。本方法给出了三个优化目标,第一个优化目标是执行时间快：能耗感知的组合方法不能以大幅度的降低执行时间来提高能耗利用率。第二个优化目标是可靠度高：组合服务中的某些原子服务可能服务率较低,但可靠性较高,因此这类原子服务也具有一定优势,它们组合在一起导致较高的总体可靠度。第三个优化目标是能耗低：根据能耗模型对服务进行选择,本文使用的M/M/c模型更贴近实际情况,能更多的减少组合服务的整体能耗。为了加快求解速度,本文用混合算法计算最优流量分布。最后在不同服务规模(抽象服务数和候选服务数不同)下进行了服务的节能效率试验,结果表明本方法在满足服务请求的前提下可以极大的降低组合服务的能耗。 3.在互联网环境中,Web服务组合具有两种不确定性,一是服务调用结果的不确定性,二是QoS的不确定性。以前的研究利用离散时间的马尔可夫决策过程对具有不确定性的服务组合建模,将服务的QoS聚合值作为即时报酬,最后求得服务组合的最优策略。这要求每个状态的转移概率必须事先已知,但是转移概率是较难得到的。此外,以前的研究没有考虑QoS的取值具有一定的概率分布。本文借鉴了相关文献提出的模型的优点,进一步扩展了服务组合模型,使用具有概率分布的QoS值对服务组合建模,本文提出的方法利用了机器学习算法获得最优的组合服务。实验结果表明本方法的学习周期短,适应能力强。 4.以前的学者为了简化研究,认为服务QoS属性间的关系是相互独立的,因此不能很好地度量候选服务对客户的价值。本文的方法考虑到服务的QoS属性具有相关性,因而对QoS权重的赋值能更准确的反映服务质量。此外,企业的业务能力与IT水平相匹配一直以来是其发展的最高优先级,SOA是推动这种匹配的关键技术。然而,如果没有一个可以衔接战略,战术和运作层面的服务组合方法,使用SOA的优势将很难在企业的业务层展现出来。考虑到上述两方面的因素,本文提出了一种改进的层次分析法用于组合服务,该方法结合战略、战术和运作层面的决策,同时考虑相互依赖的QoS属性,对服务组合的方案进行了排序,从中可以选择出最适合企业战略的组合方案。
[Abstract]:On the Internet, more and more network resources, including storage resources, computing resources, application resources and software resources, can be acquired and accessed, so that the Internet is gradually becoming an open distributed computing platform. and the service-oriented computing (SOC) and the service-oriented architecture (SOA) are the latest development directions of the current distributed computing and software development, can help enterprises or Internet users to develop applications and business processes more flexibly, quickly and at low cost, The combination of Web services is the key technology for service computing. With the development of cloud computing, e-commerce and Internet, the environment of the Web service combination has changed. In order to achieve universality, many QoS-based Web service combination methods are designed for a general-purpose environment, and some new problems and machines in a specific environment are not considered In this paper, the environment of Web service combination is subdivided, and the combination of Web services based on QoS is studied. The emphasis is to improve the traditional method to adapt them to these specific rings. The new method is put forward to make full use of the advantages of these specific environments and to further study the new problems brought by the specific environment, and put forward the corresponding solution The main contents and contribution of this paper are as follows: next:1. Since the Internet is made up of sub-networks with different delays, users at different network locations have the same service quality as the same Web service There is a difference. However, the service provider typically only provides an average of the service QoS as the service's evaluation index, resulting in the fact that most of the current service combining methods do not take into account the network transmission performance and bring the Web services to users at different locations In addition, in the case of limited capacity of Web service processing, multiple physical services are used to satisfy the service request of multiple users, not only can improve the reliability of the combined service and the ability to resist the damage, but also can remarkably reduce the total service of the Web combination service. However, the existing methods in such a case study the dynamic combination of services In this paper, by using the queuing network model, the quantitative index of the network QoS is introduced, the constraint conditions and the upper and lower limits of the optimization variables are given, and a runtime service group is proposed. the method comprises the following steps of: firstly, selecting a group of dominant entity services by using a non-linear optimization theory, The experiment shows that the method proposed in this paper is suitable for the dynamic change of the parameters in terms of the optimality and the efficiency of the implementation. 2. The previous research approach focuses on the study of atomic Web services or cluster services in terms of energy efficiency, while ignoring the entire portfolio In this paper, the energy consumption of the whole combined service is calculated according to the energy consumption model of the Web candidate service. The method provides three optimization objectives, and the first optimization goal is to perform time fast: the combination method of energy consumption perception cannot be improved with a large reduction in execution time The second optimization goal is high reliability: some of the atomic services in the combined service may have a lower service rate, but the reliability is high, so this type of atomic service also has the advantage that they are combined to lead to higher Overall reliability. The third optimization goal is low energy consumption: the M/ M/ c model used in this paper is closer to the actual situation and can reduce the combined service more by selecting the service according to the energy consumption model In order to speed up the speed of the solution, the hybrid algorithm is used to calculate the maximum energy consumption. At last, the energy-saving efficiency test of service is carried out under different service scale (the number of abstract service and the number of candidate service). The result shows that the method can reduce the combination greatly on the premise of satisfying the service request. the energy consumption of service.3. In the Internet environment, the combination of Web services has two kinds of uncertainty, one is the uncertainty of service call results, and the other is Q The uncertainty of S. The previous study uses the discrete time Markov decision process to model the service combination with uncertainty, and the QoS aggregation value of the service is used as the real-time compensation, and finally the service is obtained. The optimal strategy of the combination. This requires that the probability of transition for each state must be known in advance, but the transfer almost The rate is hard to get. In addition, previous studies have not taken into account the value of the QoS This paper, based on the advantages of the model proposed by the relevant literature, further extends the service combination model, and uses the QoS value with the probability distribution to model the service combination. The method proposed in this paper is obtained by using the machine learning algorithm. The experimental results show that the learning cycle of the method 4. The relationship between the QoS attributes of the service is independent from each other in order to simplify the study, so it is not good to measure the waiting time. The method of this paper takes into account the correlation of the QoS attributes of the service, so the assignment of the QoS weight can be more accurate reflection of the quality of service. In addition, the business capacity of the enterprise has been the highest priority for its development since it has matched the IT level, and SOA is the promotion The key technology of this match. However, if there is no one that can connect to the strategic, tactical, and operational aspects, the advantage of using SOA will be hard to be in the enterprise The business layer of the industry is presented. In view of the above two factors, this paper presents an improved analytic hierarchy process for the combination of services, which combines the decision of the strategic, tactical and operational aspects, taking into account the interdependence of the QoS attributes and the service group. The combined scheme is sorted, where you can select the most appropriate
【学位授予单位】：北京邮电大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TP393.09

【相似文献】