【作者】 何华；

【导师】 庞善臣；

【作者基本信息】 山东科技大学 ， 计算机软件与理论， 2011， 硕士

【摘要】 随着计算机技术、云计算和经济全球化的迅速发展,工作流技术被广泛应用到各种领域中。然而,所有的领域都存在某种程度和形式上变化因素,这意味着在业务过程执行过程中,组织或企业需要调整它们的活动以应对这些影响因素。为了使工作流能够适应日益激烈的竞争环境、动态多变的市场环境和个性化的用户需求,必须提高工作流的动态性、柔性和适应性。这些特性在给企业带来巨大经济效益的同时,也对工作流模型提出了巨大挑战,如经常会使过程实例在运行过程中偏离预定义的过程并失去控制。如何处理工作流实例运行过程中的跃迁,使变换后的过程既能实现原有目标,又能避免运行脱节等一系列问题,成为当前迫切需要解决的问题。工作流的动态变化主要包括业务流程变动、实例迁移、工作流演进等,本文主要研究建模动态工作流的技术,以及实现工作流实例迁移的跃迁规则。首先,本文针对具有资源约束的工作流、具有随机性的工作流分别提出了动态库所Petri网(DPPN)和变迁可编程Petri网(TPPN)。在此基础上,我们考虑具有由变量(或属性)描述的本地存储的库所,提出了可编程Petri网(PPN),详述了它的变迁发生规则,并给出了PPN与高级Petri网等价的构造性证明。然后,通过修改PPN得到了可编程工作流网(PWN)的定义,并深入分析了PWN的动态变化机制。然后,我们将网重写系统结合PWN得出了基于PWN的重写网的定义和重写规则。PWN重写网的模型和规则能够实现各种变化模式的所有操作,并且能够改变控制流变化的支持特征。其中,支持特征能够从系统的角度对比现有的过程管理技术。有两种方式可以改变重写网的业务流程：(1)改变PWN的编程函数。(2)通过改变现有业务流程的结构来改进当前系统的性能或添加、删除和更新服务以适应环境和业务需求的变化。为处理运行实例的跃迁,提出了守恒状态等的定义,并证明了某些性质。最后,针对包含动态服务组合的工作流的随机性,我们引入随机网络演算,用到达曲线和服务曲线,分析基于网络的服务组合中的时延和数据积压界限。由此可以得出随机服务保障下动态服务组合的QoS评价,并通过系统模拟与实例来分析两种情景下的性能。更多还原

【Abstract】 Along with the rapid development of computer technology, cloud computing and economic globalization, workflow technology is currently being applied in quite diverse domains. However, the factors of change are present in some degree and form in process in almost all domains. It means that organizations or enterprises need to adjust their activities to deal with these factors in the execution of business processes. In order to adapt the increasingly fierce competition, dynamic changeable market environment and personalized user requirements, the dynamics, flexibility and adaptability of workflow systems must be improved. These characteristics not only bring to enterprises huge economic efficiency, but also provide a huge challenge to workflow models. For example, process instances are usually deviating from the pre-defined processes in runtime and loosing the control. How to handle the transition of the workflow instance in runtime, guarantee the transformed process which can not only realize the original target, but also avoid running detached and so on, become an urgent problem to be solved.The dynamic changes of workflow mainly include process changes, instance migration and workflow evolution and so on. This paper mainly focus on the study of modeling technology for dynamic workflow, as well as the transition rules of realizing workflow instance migration.Firstly, this paper proposes Place-Programmed Petri Nets (PPPN) and Transition-Programmed Petri nets (TPPN) which respectively aim at workflows constrained by resources and stochastic workflows. Based on these, we consider places which have local storage described by variables (or attributes), and propose the definition of Programmed Petri Nets. This paper further describes its transition firing rule, and proves equivalence between PPPN and colored Petri nets. After that, we propose the definition of Programmed Workflow Nets (PWN) by modifying the definition of PPN, and deeply analyze dynamic mechanism of PWN.Secondly, we propose the definition of rewriting nets based on PWN and its rewriting rules. The models and rules of rewriting nets based on PWN can realize all operations of various changing modes, and can alter the support characteristics, which can compares existing process management technology from the view of systems, of the changes of control flows. There are two ways to change the business processes of rewriting nets based on PWN. One is to change the programming functions of PWN, the other is to change the structure of current business process in order to improve the performance of the system, or to add, delete and update service in order to adapt environments and the change of business requirements. For the sake of dealing the transition of instances in runtime, we propose some definitions such as conservative state, and prove some properties.Lastly, we introduce stochastic network calculus on account of the stochastic workflows including dynamic services composition. We apply arrival curves and service curves to analyze the delay bounds and backlog bounds of services composition based on communication network. It can be concluded the QoS evaluation of dynamic services composition under stochastic service guarantee. Meanwhile, we analyze the performance of two scenarios through instance simulation.更多还原