基于QoS度量的Web服务组合及容错方法研究

发布时间：2018-06-19 02:03

本文选题：服务计算 + 服务组合　；参考：《北京交通大学》2016年博士论文

【摘要】：服务计算模式允许组合各种Web服务,获得新的增值服务,实现了服务的重用。然而,由于服务数量的快速增加,Web服务的质量(Quality of Service, QoS)良莠不齐,在执行组合应用的过程中,可能由于某服务是恶意服务(携带病毒)或其QoS与服务提供商承诺的不一致而导致组合服务调用失败,也可能因环境变化而导致服务执行时所在的计算节点失效,导致组合Web服务应用不可用。因此,如何在海量的Web服务中,根据用户的服务请求,构建满足用户QoS需求的服务组合及容错方法仍然存在诸多挑战,本文针对上述背景下的若干关键问题进行了深入的研究,取得的主要成果如下：(1)研究基于QoS信任下的Web服务组合问题,提出了基于两个阶段的神经网络及QoS等级划分的服务组合方法,该方法利用两个阶段的神经网络筛选QoS受信任的Web服务,在保证全局约束条件及所有的组合成员服务都具有很好的QoS信任的条件下,建立等级划分的服务组合模型。然后提出了基于等级划分的服务组合算法,最大化组合后服务的效用值,进而获得QoS可信的近似最优解。实验结果表明,非信任服务的识别率达到90%,明显高于传统的CorrelationLens识别方法。此外,与整型规划方法相比,本文提出的基于QoS等级划分的组合算法的求解速度提升了近3倍,且获得的解已非常接近全局最优解,更好地确保组合服务的QoS指标受信任和最优。(2)研究大规模备选服务下的服务组合效率问题,提出了一种以快速抽样为核心的Web服务组合方法FAQS,该方法采用统计学抽样理论对Web服务QoS数据库进行研究,并对每个效用值区间内的服务进行抽样；其次,在样本空间内,对于所抽取的典型Web服务进行建模,即以Web服务效用值与Web服务使用频率值为目标参数建立服务选择模型；最后,根据得到的样本空间内的服务组合结果,在完全空间中进一步优化,选出满足全局QoS约束且保证QoS效用值与服务频率值最大化的服务。实验结果表明,该方法可以在大规模备选服务中准确、高效地获取服务组合结果。(3)研究组合方案中成员服务失效的问题,提出了组合方案的可靠性模型FTDes,即决策模型与优化模型,旨在提升Web服务组合方案的可靠性。在决策模型构建过程中,通过选取决策参数、构建决策矩阵、设计决策模型,为失效的成员服务找到可替代的备选Web服务集合；在优化过程中,把策略优化选择问题归结为0-1整型规划问题,提出顶点凸壳的过滤方法并结合标准的整型规划器CPLEX得到优化的服务替换方案。实验结果表明,与传统的决策方法(包括层次分析法APH、证据推理法ER、喜好值排序技术的决策模型TOPSIS、线性分配法LAM)相比,本文提出的方法从决策准确度、响应时间方面要优于传统方法,与此同时,本文分析了优化选择过程、TopK算法以及不同参数对于决策结果的影响,更好地保证组合服务的整体QoS指标得到进一步维持与优化,有效地提高了Web组合方案的可靠性。(4)研究组合Web服务应用的节点失效问题,提出了环形心跳检测机制与轻量级恢复机制,旨在提升组合Web应用的可靠性。在环形心跳检测机制中,为了减轻中心节点的压力,基于环形拓扑的检测方式设计了分布式的心跳算法,并在每个节点上部署该心跳算法；针对传统恢复方法时间开销过大的问题,提出了轻量级的恢复机制,设计了考虑负载重新分配的恢复算法,同时考虑集群负载超出阈值时,在环形结构中增加新的节点,从而降低集群负载,避免因负载过高而引起节点失效。实验结果表明,与传统的日志分析检测方法和系统级恢复方法相比,该检测与恢复机制的开销小、延迟少,避免了服务器节点失效导致组合Web服务应用的整体QoS降低,提升了组合Web服务应用的可靠性。综上所述,本文对QoS度量的Web服务组合中的组合方法与容错方法展开深入研究,提出了等级划分与快速抽样的Web服务组合方法,以及环形心跳检测机制与轻量级恢复机制,提升了Web服务组合的效率、准确性与可靠性,具有一定的理论意义及广泛的应用价值。
[Abstract]:The service computing mode allows the combination of various Web services to obtain new value-added services and achieve service reuse. However, because of the rapid increase in the number of services, the quality of the Web service (Quality of Service, QoS) is intermingled, and in the process of implementing the combined application, it may be from a service is a malicious service (carrying a virus) or its QoS and service. The inconsistent commitment of the supplier leads to the failure of the combined service call. It may also cause the failure of the computing node in the service execution due to the environment change, which leads to the unavailability of the combined Web service application. Therefore, how to build a service composition and fault tolerance method to meet the user's QoS requirements in the mass Web service There are many challenges in this paper. The main achievements of this paper are as follows: (1) research on Web service composition based on QoS trust, and propose a method of service combination based on two stages of neural network and QoS hierarchy. This method uses two stages of neural network. The collaterals selected QoS trusted Web services to establish a hierarchical service portfolio model under the condition of guaranteeing global constraints and all the combined member services having good QoS trust. Then, a hierarchy based service composition algorithm was proposed to maximize the utility value of the combined service, and then to obtain the approximate optimal QoS reliability. The experimental results show that the recognition rate of the untrusted service reaches 90%, which is obviously higher than the traditional CorrelationLens recognition method. In addition, compared with the integer programming method, the solution speed of the combined algorithm based on the QoS hierarchy is up to nearly 3 times, and the obtained solution is very close to the global optimal solution, and it is better to ensure the combination clothing. The QoS index of service is trusted and optimal. (2) to study the efficiency of service composition under large-scale alternative service, a Web service combination method, FAQS, which is based on rapid sampling, is proposed. The method uses statistical sampling theory to study Web service QoS database and sample each service value interval. Secondly, In the sample space, the typical Web service is modeled, that is, the service selection model is set up with the Web service utility value and the Web service use frequency value as the target parameter. Finally, according to the result of the service composition in the sample space, it is further optimized in the complete space, and the global QoS constraint is satisfied and the QoS utility value is guaranteed. The experimental results show that the method can accurately and efficiently obtain the service composition results in the large-scale alternative service. (3) the problem of member service failure in the combination scheme is studied, and the reliability model FTDes of the combination scheme, that is, the decision model and the optimization model, is designed to improve the Web service composition party. In the process of decision-making model construction, by selecting decision parameters, building decision matrix, designing decision model, finding alternative alternative Web service sets for invalid member service. In the process of optimization, the problem of strategy optimization is reduced to 0-1 integer rule problem, and the filtering method of vertex convex hull is proposed and combined with the standard. The experimental results show that the method proposed in this paper is superior to the traditional method in terms of decision accuracy and response time compared with traditional decision making methods (including analytic hierarchy process APH, evidence reasoning method ER, decision model TOPSIS of preference value sorting technique and linear allocation method LAM), and the experimental results show that the method proposed in this paper is superior to the traditional method in terms of decision accuracy and response time. At the same time, this paper analyzes the optimization selection process, the TopK algorithm and the influence of different parameters on the decision results, to better ensure the overall QoS index of the combination service to be further maintained and optimized, and effectively improve the reliability of the Web combination scheme. (4) study the node failure of the combination of Web service applications, and propose a ring heartbeat detection. The measurement mechanism and lightweight recovery mechanism are designed to improve the reliability of the combined Web application. In order to reduce the pressure of the central node in the ring heartbeat detection mechanism, a distributed heartbeat algorithm is designed based on the loop topology detection method, and the heartbeat method is deployed at the top of each node, and the time overhead of the traditional recovery method is too large. A lightweight recovery mechanism is proposed, and a recovery algorithm considering load redistribution is designed. At the same time, new nodes are added to the ring structure when the cluster load exceeds the threshold, thus reducing the load of the cluster and avoiding the node failure due to the high load. The experimental results show that the traditional log analysis method and system level are shown. Compared with the recovery method, the detection and recovery mechanism has less overhead and less delay, which avoids the failure of the server nodes to reduce the overall QoS of the combined Web service application and improves the reliability of the combined Web service application. The Web service combination method of level division and rapid sampling, as well as the ring heartbeat detection mechanism and lightweight recovery mechanism, improve the efficiency, accuracy and reliability of the Web service composition, and have a certain theoretical significance and wide application value.
【学位授予单位】：北京交通大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TP393.09

【相似文献】