【作者】 龙浩；

【导师】 邸瑞华；

【作者基本信息】 北京工业大学 ， 计算机应用技术， 2010， 博士

【摘要】 利用网格技术将异构及地理分布的自动化设施和研究人员聚集起来,成为一个协同的研究环境,是目前科学研究和商业应用的主要模式。网格工作流可以方便地构建、执行、管理和监控网格应用,使得网格应用自动实施并且效率较高,能够很好地支持协同工作和资源共享的需要。网格范围的扩大和资源的不断增加,给系统提出了智能化服务查找和服务组合的需求,传统的网格工作流不再适用。本文将语义Web技术和工作流技术引入到网格中,提出了“网格语义工作流”的概念,并给出了通用的网格语义工作流体系结构,对语义服务的自动标注和语义服务分类、网格工作流本体建模、工作流自动组合和优化调度进行了研究,主要内容包括:(1)网格语义工作流体系结构建立通用的网格语义工作流体系结构是本文研究的第一步。首先提出了网格语义工作流体系结构的概念,针对工作流与底层资源关系的疏密程度,提出了三层工作流抽象层次,分别是抽象工作流、具体工作流和可执行工作流。基于现有的工作流技术、网格计算、语义Web的相关规范和技术基础之上,设计了一个通用的网格语义工作流体系结构。研究了在本体库的支持下,进行各类语义资源的管理,包括资源语义标注、语义资源的分类和语义服务的Qos计算。(2)网格工作流本体建模本体是实现知识表示的重要方法,是对网格资源进行语义化描述的基础。本文针对协同应用的特定需要,通过研究工作流元模型,针对网格工作流涉及的知识领域和相关的概念术语,研究支持对组织、资源和过程进行语义建模的统一本体框架。基于特征知识建立应用领域的本体知识模型。并以此作为基于语义的网格应用的基础;研究网格内组织、人员和安全控制等概念术语,建立网格组织本体模型;研究OWL-S本体模型,结合实际需要,对其进行扩展,加上描述服务和资源Qos的语义本体,建立网格服务本体模型。(3)基于Qos感知的语义工作流自动生成工作流的自动生成,是一个非常有挑战性的难题。语义化描述的网格服务,为智能化的工作流自动生成,提供了基础;网格的分布性、异构性、自治性和动态变化的特点,对工作流的正确性、安全性、可靠性等都有很大影响。网格语义工作流自动组合必须同时考虑用户的IOPR功能需求和Qos性能约束。利用语义服务的相关向量及在此之上定义的一组运算来描述服务组合,利用服务组合匹配度来衡量服务组合满足用户需求的程度,将Qos感知的网格语义工作流自动组合问题转化为线性规划问题;对人工鱼群算法进行改进,加入遗传算法中的交叉算子和变异算子,研究适用于解决语义工作流自动组合问题的改进人工鱼群算法IAFSSCM(Improved Artificial-Fish-School-Based Service Composition Method);用Petri网来形式化描述服务组合方案,并研究支持控制结构的工作流Qos估算方法。与其他的AI算法比较,改进人工鱼群算法能够很快地找到满足服务组合匹配度的多个组合方案;利用petri网来计算工作流Qos,能够体现数据交换中隐含的控制结构。仿真实验证明,算法能够很好地支持Qos感知的网格语义工作流的自动生成,并能保证所生成的工作流的适应性和可靠性。(4)网格工作流调度在网格应用中,时间费用优化问题是用户最为关心的问题。采用相对效费比来描述调整服务对工作流完工期和总费用的影响,并以此为基础给出给定截至期的工作流的时间费用优化算法---RTCR算法;对于复杂结构的工作流,分析工作流的结构,考虑对工作流进行任务分段以降低问题复杂度,得到分段调度算法SL。SL算法将全局优化问题转化为几个局部问题,对于段内采用RTCR算法进行优化。通过理论分析和仿真试验,该算法可以得到较好的优化性能。(5)结构工程语义网格工作流管理系统原型在以上研究的基础上,建立了一个应用于结构工程实验领域的语义网格工作流管理系统原型,完成了一个模拟结构工程试验从定义、验证到运行的过程。更多还原

【Abstract】 It’s the orientation of science research and business to congregate geographically distributed and heterogenous facilities and researchers as a virtual environment by using grid technology. Coordination and resource sharing are the main purposes of building grid systems, which can be partly realized with grid workflow. A workflow is a series of associated tasks linked together, which can be easily constructed, managed, monitored, and can execute automatically and efficiently in case submitted. Expanding scale and increasing resources induce intelligent resource inquiry and service composition as essential requirements, which are out of the ability of the conventional grid workflow. The thesis inducts semantic web and workflow technologies into grid, gives out the common concept of grid semantic workflow, constitutes the ontology-based workflow architecture, and studies annotation of semantic services, classification of semantic services, construction of grid workflow ontology model, automatic workflow composition, and time-cost optimization scheduling. The detail is as follows:Grid semantic workflow architecture. By analyzing the characteristics of semantic grid applications, the concept“grid semantic workflow”and a common architecture are given. According to the relation with lower resources, grid semantic workflow can be fractionized as abstract workflow, concrete workflow and executable workflow. Based on exsiting technologies and specifications of workflow, grid, semantic web etc., the architecture is constituted as a five layers structure composing of user layer, semantic workflow engine layer, semantic layer, web service layer and physical resource layer. The thesis lay emphasis on grid semantic resource management, including the annotation and classification mechanics of semantic services, and the weighing of Qos parameters for semantic services.Grid workflow ontology model. Ontology is an important expression of domain knowledge, and is the base to semantically describe resources. Aiming to the special demands of coordination and cooperation of grid application, the ontology framework is presented, which composes of domain ontology, organization ontology and service ontology, and OWL is used to define ontology concepts.Automatic creation of Qos-awared grid semantic workflow. Under dynamic grid environment, many uncertain factors affect the correctness, security and reliability of grid workflow strongly. A service composition algorithm is proposed, which considers user’s IOPR requirements and Qos constraints jointly. Correlation vector of semantic services and a set of operations above it are used to represent service composition, and service composition is transferred as linear programming problem. An algorithm called IAFSSCM(Improved Artificial-Fish-School-Based Service Composition Method) is presented to solve service composition. A Qos weighing algorithm is introduced to evaluate user’s Qos requirements, which uses Petri net as formalization tool. Simulation result demonstrates the method’s affectivity.Semantic workflow scheduling. Aiming at the semantic Workflow scheduling with the objective of time-cost optimization, workflow is divided into segments according to its structure, and the deadline is partitioned to different segments. For tasks in each segment, their tasks are scheduled with Relative Time-Cost Rate algorithm. Simulative experiments show its performance improvement compared with other AI method.Grid workflow prototype and experiments. The prototype development is followed the architecture and research work introduced above. A simple simulated structural engineering experiment is demonstrated.更多还原