Optimization Methods For Cloud Service Composition And Scheduling In Logistics Domain

The rapid development of modern service industry requires the facilitation of socio-technological complex service systems. Bi-lateral resource integration service pattern(BIRIS) is a dominant pattern for current service systems, where massive customers and massive service providers are aggregated together via a broker service platform, the service relationships are established between both sides, and the service execution process is provisioned in a centralized way. One of the most important issues in BIRIS is to select and organze available candidate services to effectively build the high reliable service systems which satisfy personalized customer requirements.However, the multi-party participation, the concentration and complexity of interactions between them, the high personalization of customer requirements, and various uncertainties in external environment, all bring great challenges to the development and execution of service systems, such as(1) the higher personalization of customer requirements require higher diversity on service functionalities and performances;(2) the complex collaborations between multiple providers and customers lead to a lot of uncertainties and dynamics on environment, resources, ad requirements;(3) multiple parties expect to obtain maximum benefit not only from current service but also the potential and continuous benefit from future services, and there are contraditions between the expectatitons of different parties.To solve above issues, this dissertation uses ocean transporation service as the background to present the idea of BIRIS-based domain service cloud. Virtual Service Resource(VSR) is employed as a thread running through the full lifecycle of service systems. The research focuses on two key phases, i.e., service system development and execution. The corresponding models and algorithms are presented to solve the key issues including domain service cloud architecture, resource virtualiation, service composition, and service scheduling. The objective is to make the service systems have the capacity of fulfilling personalized requirements in a high-efficiency and cost-effective way.There are three scientific problems that this dissertation tries to solve: domain-specific service cloud architecture, resource virtualization and the performance evaluation; domain-specific service cloud resource composition in development phase, with the objective of improving composition efficiency and the satisfication degree of personalized requirements; domain-specific service cloud resource dynamic scheduling, to deal with the uncertainties.Detailed research works include:(1) Based on the primary concepts of service system, service value network, and bi-lateral resource integration service pattern(BIRIS), we present the architecture and internal resource organization pattern of BIRIS-based domain service cloud, with the resource virtualization method in cloud service environment. A special type of Virtualized Service Resource(VSR), service pattern, is particllarly studied. The performance indicators of service resource virtualization and the value-driven service value network design method for cloud services are presented, too.(2) Based on the summarization of customer personalization features, we present the Baysian Brief Network based service personalization identification method, to establish the probabilistic relationships between customer personalization features and the corresponding service features. Aiming at behavior-type service resources, we present a pattern-based service composition algorithm where greedy covering policy and artificial bee col ony algorithm are employed to solve the optimization problem, so as to achieve the maximum composition efficiency and the minimum composition cost. Aiming at product-type service resources, we present a dynamic pruning based resource integration method. Considering the preference on risk, we present an insurance-based service reliability enhancement method, with the objective of making service composition solutions have higher anti-risk capacity.(3) Aiming at various uncertainties during service execution, we focus on the problem of dynamic service scheduling with the objective of fulfilling service execution with minimum scheduling cost. Two strategies are adopted respectively, i.e.,(a) analyze the event flow generated during service execution to predicat e beforehead the potential risks that might exist in those un-executed service activities, and then specific scheduling actions are adopted to avoid them;(b) after one uncertainty event actually happens, specific sechduling actions are identified to make the service exection return to the predetermined execution plan by minimum cost. k-neighboring algorithm and Markov Decision Process(MDP) are adopted for the two strategies, respectively.(4) Based on Software as a Service(Saa S) and Cloud computing techn iques, a BIRIS-based ocean transportation service cloud platform is designed and developed, with a set of infrastructural services facilitating service personalization, service monitoring and scheduling, effective service delivery, etc. A set of typical ve rtical and horizontal service systems composing the business cloud of the platform are implemented. The platform has been applied in real-world business, and the feasibility and effectiveness part of the proposed theoretical methods are partially validated through the application of this platform.Research work in this paper is supported the Natural Science Foundation of China(NSFC), the National Science & Technology Pillar Program, and the Science and Technology Development Program of Shandong Province.