面向SOA的服务调度算法及应用研究

发布时间：2018-04-09 23:35

本文选题：面向服务的体系架构　切入点：Web服务　出处：《哈尔滨工程大学》2016年硕士论文

【摘要】：在企业信息化转型过程中,由于各管理系统往往是分阶段逐步改造或创建的,在架构、接口等方面可能存在差异性,进而导致难以集成协作的问题,为解决该问题,通常会选择SOA架构进行系统集成,不仅能够屏蔽系统间接口和协议上的差异,还能够更方便的在系统间实现业务代码复用。选择SOA架构意味着各系统会通过Web服务进行通讯和交互,在整体系统的拓扑结构中,处于中心位置的子系统很可能出现热点服务,继而面临请求规模超出服务处理能力的困境,此时需要考虑对请求进行合理调度。请求调度问题在不同场景下,有不同的研究重点。为了解决服务处理能力不足的问题,一种思路是冗余部署,通过扩展服务提供者数量达到提升服务处理能力的目的。在该场景下,请求调度问题研究的是如何在多个服务提供者中合理选择转发目标,实现负载均衡。另一种思路是保持单一的服务提供方,通过考察请求的特征,优化请求的响应顺序,达到提升服务整体处理能力的目的。本文对请求调度问题的研究就是在采用后一种解决思路的场景下,希望通过提出合理的调度策略,能够在不增加额外的部署及管理成本的前提下,改善服务调用请求的成功率,提高服务的吞吐能力。本文首先对问题进行了分析,通过比较默认调度策略和其他改进型调度策略的优缺点,确定了本文算法的研究方向。针对SOA的集成特点,提出了服务的获益策略并制定了该策略下的获益指标集合,并基于对请求历史的分析,实现对请求结果有效性的预估。最终将获益值、有效性和急迫度的带权计算结果作为请求调度的依据。通过对比实验,在不同请求场景下验证了算法在调度效果上的改进。最后在实际集成工具中应用了本算法,实践了算法的应用场景。算法的不足之处在于调度效果在一定程度上依赖于服务在不同参数下结果计算时间的差异程度,这使得算法更适用于复杂业务服务。
[Abstract]:In the process of enterprise information transformation, because every management system is often transformed or created step by step, there may be differences in structure, interface and so on, which leads to the problem of difficult integration and cooperation, in order to solve this problem.SOA architecture is usually chosen for system integration, which can not only shield the differences in interfaces and protocols between systems, but also implement business code reuse more conveniently among systems.Choosing SOA architecture means that each system communicates and interacts with each other through Web services. In the topology of the whole system, hot services are likely to appear in the central subsystem, and then the request scale exceeds the processing capacity of the service.At this point, we need to consider the reasonable scheduling of the request.The request scheduling problem has different research emphasis in different scenarios.In order to solve the problem of insufficient service processing capacity, one way of thinking is to expand the number of service providers to enhance the service processing capacity.In this scenario, the problem of request scheduling focuses on how to reasonably select forwarding targets among multiple service providers to achieve load balancing.Another idea is to maintain a single service provider and optimize the response order of the request by investigating the characteristics of the request to achieve the purpose of enhancing the overall processing capacity of the service.In this paper, the research of request scheduling problem is to improve the success rate of service call request without adding additional deployment and management cost by proposing a reasonable scheduling strategy under the scenario of adopting the latter solution.Improve the throughput capacity of the service.Firstly, this paper analyzes the problem, compares the advantages and disadvantages between the default scheduling policy and other improved scheduling policies, and determines the research direction of the algorithm in this paper.According to the integration characteristics of SOA, a service benefit strategy is proposed and the benefit index set under the strategy is formulated. Based on the analysis of the request history, the validity of the request result is estimated.Finally, the weighted calculation results of benefit, validity and urgency are used as the basis of request scheduling.Through the contrast experiment, the improvement of the scheduling effect is verified in different request scenarios.Finally, the algorithm is applied in the practical integration tool, and the application scene of the algorithm is put into practice.The disadvantage of the algorithm is that the scheduling effect depends to some extent on the difference of the computing time of the service under different parameters, which makes the algorithm more suitable for complex business services.
【学位授予单位】：哈尔滨工程大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP393.09;TP301.6

【相似文献】