Research On QoS-aware Services Composition In Internet Environment

Service-Oriented Computing (SOC) and Service-Oriented Architecture (SOA) are new research directions in distributed computing and software development, which can help the enterprise to build applications and business processes more rapidly, cheaply and flexibly. Web service composition is the key underlying technology. In order to keep up with the fierce market competition and meet the users' strict requirements, web services for the composition should be selected based on QoS and monitored continuously during runtime to keep or improve the QoS of web service composition. The process is called QoS-aware services composition.With the rapid development of Internet and e-commerce technologies, more and more web service compositions occur in Internet environment. Most of the current approaches for QoS-aware services composition are designed for generic environment, which don't consider the specific requirements and challenges in Internet environment. In this thesis, we research on the problem of QoS-aware services composition in Internet environment and focus on the following three aspects:adapt traditional approaches to Internet environment, propose new approaches to make full use of the advantages in Internet environment, research on the new issues in Internet environment and provide corresponding solutions. Major work and contributions include:First, we research on the basic issues faced when applying traditional approach to the problem of QoS-aware services composition in Internet environment and propose a new optimizing process. We analyze new issues of web service composition in Internet environment, discuss QoS model of web service in the environment, investigate the influence of the services distribution and execution engine deployment location on the performance of composite web service, and propose an integer linear programming based optimization process which improves the traditional approach from the perspective of performance and security. Experiments show the effectiveness of the new approach and the performance is also acceptable.Second, we propose the QoS-based multi-granularity services composition problem and its solution. In order to fully utilize the abundant web services resources in the Internet, we propose and formally define the problem of QoS-based multi-granularity services composition (QMGSCP). By introducing the "Multi-Granularity Constraints", QMGSCP is formulated as a mixed integer linear programming problem. In this way, better composition solution can be achieved since more web services will be available as candidate. Experiments show the improvement in performance, optimization rate and success rate.Third, we research on QoS-guaranteed web service composition management problem and propose a QoS-based, bottom-up services replacement approach for web service composition. In order to adapt to the dynamically changing Web service resources in open environment and ensure that the composite web service meet the business requirements continuously, we propose a QoS-based, bottom-up services replacement approach. Experiments show that the proposed approach could find more candidate Web services for replacement thus maintain or improve the QoS of the composite Web service more effectively.Fourth, we research on the reliability correlation of component web services in Internet environment, propose a tree-based reliability model and improve the services selection process. We categorize common-cause failures (CCFs) of composite web service in Internet environment and propose a tree-based reliability model that considers the correlation of component services. Based on the model, we research on the CCF-aware QoS-based web service selection process and propose two algorithms (POST-CCG and PRE-CCG) to avoid or mitigate the impact of CCFs, thus to ensure that reliability of the web service composition meet the requirement.Last, we design and implement the experimental system AQSCS (Advanced Quality-aware Services Composition System). AQSCS provides simulation environment and implements the algorithms for QoS-aware services composition in Internet environment proposed in the thesis. The system is extensible and can be used as a basic platform for further research.