【作者】 马满福；

【导师】 胡正国；

【作者基本信息】 西北工业大学 ， 计算机科学与技术， 2007， 博士

【摘要】 使用经济理论和方法解决网格资源调度问题，是当前网格研究的重点和热点。其内容涉及基本的经济模型(主要是拍卖和自由竞争)、信任问题、安全问题、监测、控制、记账和支付等几个方面。就当前的研究情况，从内容涵盖的范围看，还不能满足一个独立运行的基本经济要素，如评价、稳定性控制、契约等就没有涉及；从研究的深度看，研究内容还停留在基本的原理性试验阶段，不能满足实际需要，也不够“经济化”。因此，有更多的研究和试验工作尚有待于进一步深入、全面的开展。本文以自由竞争模型为基础，对其中的信任、价格、资源调度中的优化等问题进行了深入的探讨，主要研究工作包括以下几个方面：1．提出了受约束的竞争经济模型(Restricted Commodity Market Model，RCMM)。该模型基于自由竞争理论，以实体间相互信任为基础、以时间和预算来约束资源调度，从而保证调度的有效性和稳定性。针对该模型，以多代理系统(Multi Agent System，MAS)描述其组成结构。2．提出了以独立第三方的方式对实体，即资源、网格服务消费者(Grid Services Consumer，GSC)和网格服务提供者(Grid Services Provider，GSP)，进行可信任评价的方法，其评价结果形成实体在RCMM中的可信任基础，最终构成选择资源的依据。其内容包括资源信任度、GSC和GSP信誉度的概念和相应的计算方法；信誉度评价算法；以信誉度为依据的资源选择算法；评价体系结构；等等。3．提出了以Ramsey定价为基础的最高限价策略来实现价格管制，以保证价格的平稳性。在价格协商中，实现了RCMM中的T(?)tonnement过程，加速协商，减小开销。4．针对资源选择优化，定义了网格距离及其计算方法，提出了以网格距离为依据的优化算法；描述了局部性特征，给出了局部调度算法：设计了资源绑定的结构及绑定描述。以此为基础，提出了绑定调度算法；结合上述三种优化思想，制定了RCMM中的优化策略。5．在RCMM中，提出了契约模型，实现了以契约为基础的资源调度管理，以支撑RCMM的经济特征。6．提出了RCMM的监测结构和评价方法。它不仅能按照传统方法采集资源所在系统中的RUR信息，而且可全面采集任务队列、资源目录等基础设施的信息，因而在为性能评估、错误发现、恢复、容错、以及预测等提供基本支持的同时，还能为市场中的价格、供求关系、信誉度评价等提供数据。更多还原

【Abstract】 Employing economic approaches to tackle resource scheduling is a hot spot inthe grid technology, which involves various contents including economic model(such as auction and free competition), trust, security, monitoring, control, tally,payment, and so on. Current though, the state of art in the grid resource schedulingis far behind other grid research areas. No economic element that supports multiplecriteria including evaluation, stable control, contract and so forth is proposed in therelated works. And most results achieved in the related projects are still atexperiment stage, which can not satisfy the practical requirements. Trying to solvethese intractable problems needs, this dissertation adopts the free competition modelas the basis, discusses the resource scheduling problems including trust, price andscheduling optimization in detail. The main contributions of this dissertation are asfollows:1. The theoretical hypothesis of the RCMM(Restricted Commodity MarketModel) model is proposed. Based on the free competition model and trustbetween entities, the resource scheduling according to the time andbudget is constrained in the RCMM model. And the architecture of RCMMmodel is composed by the multiple agent; systems.2. Trust evaluation of the entities that contain resource, GSC andGSP is performed by the independent third party organization. Accordingto this approach, the trust basis between entities and the selectioncriteria of resource scheduling are formed, including concepts ofresource trust, GSC(Grid Services Consumer) credit, GSP(Grid ServicesProvider) credit and their calculation formulas, credit evaluationalgorithm, resource selection algorithm based on credit, evaluationarchitecture.3. T(?)tonnement price negotiation process is implemented in the RCMMmodel. And a highest price fixing approach based on the Ramsey pricetheory to carry out price controlling is firstly presented, which guarantees the stability of price.4. Aiming at optimization of resource selection, the concept of griddistance and its computation method is defined, then the optimizationalgorithm based on grid distance is introduced. The local schedulingalgorithm is presented according to the description of localcharacteristics. A binding scheduling algorithm is given according tothe definition and structure of resource binding. The whole optimizationstrategy is composed by the three optimization approaches above.5. The contract-based resource scheduling management that definesthe structure and transition process of the contracts is proposed in theRCMM model, which underpins the economic features of the RCMM model.6. The monitoring and evaluation system of the RCMM model is implemented.Information, not only RUR data of the sub-system where the resource reside, butalso infrastructure data including task queue and resource directory, can be gatheredby this system. Based on the information, performance evaluation, defect detection,recovery, fault-tolerant and prediction can be supported, besides data of marketprice, supply and demand relation, credit evaluation can be provided.更多还原