云服务平台中的服务协商算法及交易过程管理研究

发布时间：2018-05-15 11:20

本文选题：多属性协商 + 贝叶斯学习　；参考：《山东大学》2014年硕士论文

【摘要】：近年来,云计算模式的势头愈演愈烈,其理念在制造业逐步兴起,很多计算机服务中心,把资源虚拟化为服务,并集中起来建立云服务平台。云制造的概念也应运而生。大量服务的聚集在更好地满足了服务需求者的各种需要的同时,也给服务分配带来了巨大挑战。作为服务分配中的一个重要问题,服务交易成为一个重要的研究课题。鉴于服务购买者的多样化需求以及自己出售服务的迫切需要,服务提供者需要以各种方式展示并销售自己的服务,因此,建立多种交易模式是服务提供者也是云平台交易系统的基本需要。目前主要的交易模式有固定价格、协商、拍卖等,相较于协商而言,固定价格和拍卖交易的研究成果已经比较成熟。而制造业和服务业目前正呈现出一种日益精细化和专业化的趋势,“按需定制”和“按需使用”的理念也越来越深入人心,服务购买者的需求表现在多个属性上。专业化服务制造链的建立,也使得服务提供者之间的服务交易在多个议题上存在异议。因此,在开放和动态的云服务平台上,多属性协商机制是解决服务交易问题的最好选择,也因此而成为很多领域的重要研究对象。不同于单属性协商(即传统的价格协商)中协商双方最终只能实现“单赢”,多属性协商是可以获得“双赢”的。双赢不仅有利于单个协商实例中交易合约的达成,更能促使协商参与方之间的长期合作,实现服务提供者的长久利益。双赢的实现主要是因为协商参与方在多个属性的不同偏好,因此在协商中获取对方偏好信息十分重要。因此,本文首先提出了一种基于贝叶斯学习机制的双赢多属性协商策略,通过学习对方的偏好信息,来指导协商策略达成双赢。同时,为了提高协商策略的有效性,本文也对学习算法做出了两点改进：第一,本文提出了双向假设空间的概念,用于改善对方代理协商行为出错时的协商结果；第二,本文还提出了一种基于协商属性让步先后序列的假设概率更新方法,使代理不必关心对方的保留效用值和让步率,从而在很大程度上提高学习结果的精度。针对多种协商模式的实现问题,本文提出了一种基于多模式的交易系统过程管理方案。目前虽然很多人在交易模式领域展开了研究,但交易系统的真正实现和应用仍然发展缓慢,且多数原型集中在单模式交易系统,这与交易过程管理方面的考虑不足有很大关系。因此我们提供了一个基于云平台服务交易系统的多模式交易过程管理架构,并给出了交易过程的流程和管理方法。另外,我们对每个交易模式采用web service和BPEL来实现,来实现单模式流程的平台独立性、灵活性、可实现性等特点。
[Abstract]:In recent years, the cloud computing model is becoming more and more powerful, and its concept is gradually rising in the manufacturing industry. Many computer service centers virtualize resources as services, and focus on building cloud service platform. The concept of cloud manufacturing came into being. The agglomeration of a large number of services not only better meets the needs of service demanders, but also brings great challenges to service distribution. As an important issue in service distribution, service transaction has become an important research topic. Given the diverse needs of service buyers and the urgent need to sell their own services, service providers need to showcase and sell their services in a variety of ways, The establishment of a variety of transaction models is the basic needs of service providers and cloud platform trading systems. At present, the main trading models are fixed price, negotiation, auction and so on. Compared with negotiation, the research results of fixed price and auction transaction have been relatively mature. At present, manufacturing and service industries are showing a trend of refinement and specialization. The concepts of "on demand customization" and "on demand" are becoming more and more popular, and the demand of service buyers is manifested in a number of attributes. The establishment of specialized service manufacturing chain also makes the service transaction between service providers different on many issues. Therefore, on the open and dynamic cloud service platform, the multi-attribute negotiation mechanism is the best choice to solve the problem of service transaction, and therefore becomes an important research object in many fields. Different from single attribute negotiation (that is, traditional price negotiation), negotiation can only achieve "single win", and multi-attribute negotiation can achieve "win-win". The win-win situation is not only conducive to the conclusion of the transaction contract in a single negotiation instance, but also to the long-term cooperation between the negotiation participants and the realization of the long-term benefit of the service provider. The realization of win-win situation is mainly due to the different preferences of negotiation participants in multiple attributes, so it is very important to obtain the preference information of each other in negotiation. Therefore, this paper first proposes a win-win multi-attribute negotiation strategy based on Bayesian learning mechanism, which can guide the negotiation strategy to achieve win-win by learning the preference information of the other party. At the same time, in order to improve the effectiveness of negotiation strategy, this paper also makes two improvements to the learning algorithm: first, this paper proposes the concept of bidirectional hypothesis space, which is used to improve the negotiation results when the negotiation behavior of the other agent is wrong. This paper also proposes a method of updating the hypothesis probability based on the negotiation attribute concession sequence, which makes the agent do not have to care about the retention utility value and concession rate of the other side, so as to improve the accuracy of the learning results to a great extent. In this paper, a process management scheme based on multi-mode transaction system is proposed to solve the problem of multi-negotiation mode. Although many people have carried out research in the field of transaction model, the real realization and application of trading system is still slow, and most prototypes are concentrated in single-mode trading system, which is related to the insufficient consideration of transaction process management. Therefore, we provide a multi-mode transaction process management architecture based on cloud platform service transaction system, and give the process and management method of transaction process. In addition, we use web service and BPEL to realize the platform independence, flexibility and realizability of single mode flow.
【学位授予单位】：山东大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.09

【参考文献】