Automatic quality of service adaptation for composite web services

• Main Author: Qiao, Ming
• Format: Others
• Published: 2009
• Online Access: http://spectrum.library.concordia.ca/976224/1/MR63228.pdf; Qiao, Ming <http://spectrum.library.concordia.ca/view/creators/Qiao=3AMing=3A=3A.html> (2009) Automatic quality of service adaptation for composite web services. Masters thesis, Concordia University.

Quality of Services (QoS) management has become an important issue for Web services. Indeed, QoS is becoming a crucial and a distinguishing criterion among functionally equivalent Web services. QoS Management consists of two complementary tasks: monitoring and adaptation. Both are very challenging because of the unpredictable and dynamic nature of Web service composition. We are motivated to solve the QoS problem by taking advantage of some characteristics of composite Web services, such as their similarity to traditional workflows. In this thesis, we propose a broker based architecture that enables dynamic QoS monitoring and adaptation for composite Web services. Our approach consists of dynamically changing the execution paths of composed Web services by instrumenting the BPEL process. A new construct flexPath is introduced for supporting alternate execution paths definition in BPEL. We developed a BPEL compiler allowing automatic instrumentation for BPEL definition files. The BPEL process is deployed using the instrumented definition files in order to interact with the QoS broker during execution. The QoS broker is a key component in our architecture and is responsible of monitoring the QoS and managing the adaptation. We propose a broker that enables runtime monitoring of QoS, prediction of potential QoS violation, and the selection of the best execution path of the process in order to improve QoS when needed. We developed a prototype to evaluate our proposed architecture. A case study is also presented through an example BPEL process and a number of partner Web services. The performance of the QoS adaptation has been analyzed and the results showed that the QoS of the BPEL process has been considerably adapted and improved comparing to the original one. In addition, we analyzed the major factors that affect the performance of our prototype tool.