|Thesis abstract: |
In Service-Oriented Computing (SOC), Service Level Agreements (SLAs) are mutual contracts between a provider and consumer of a service that define quality guarantees. Quality of Service (QoS) once de fined in such contracts may change during the life-cycle of Service-Based Applications (SBAs) due to various reasons. Accordingly, Web services need to be able to adapt dynamically to respond to such changes. With this regard, Web service adaptation and evolution are receiving huge interest in the Service-Oriented Architecture (SOA) community due to dynamic and volatile Web service environments. This thesis investigates issues of QoS contract formation between service providers and consumers and how either party can evolve independently from each other without violating the agreed contract. Acceptable changes are defi ned using a compatibility mechanism. The thesis proposes an approach for service adaptation through defi ning a fl exible QoS property description using fuzzy parameters. Partial satisfaction of quality parameters is allowed through the defi nition of linguistic variables. QoS satisfaction degree is measured using membership functions provided for each parameter. Recently, many adaptation strategies have been proposed in the literature. However, there is no overall consensus in selecting the best strategy and often adaptation requirements are not taken into account. An essential issue to be addressed is how to e fficiently select an adaptation strategy when there are different alternative strategies. Moreover, formulating QoS parameters and their relationship with adaptation behaviour of a SBA is a di fficult and challenging issue. The thesis proposes a Fuzzy Inference System (FIS) for measuring an overall QoS and selecting adaptation strategies using fuzzy rules. The overall QoS is inferred by QoS parameters, while selection of adaptation strategies is inferred by the overall QoS and other adaptation requirements such as importance of QoS and cost of service substitution. Penalties are assigned according to the degree of QoS violation for compensation purposes. Experimental results show the effectiveness of the proposed fuzzy approach for decision making in selecting between alternative adaptation strategies.