A Probabilistic Approach to Service Selection with Conditional Contracts and Usage Patterns

Service selection is a central challenge in the context of a Service Oriented Architecture. Once functionally sufficient services have been selected, a further selection based on non-functional properties (NFPs) becomes essential in meeting the user’s requirements and preferences. However, current descriptions of NFPs and approaches to NFPaware selection lack the ability to handle the variability of NFPs, that stems from the complex nature of real-world business scenarios. Therefore, we propose a probabilistic approach to service selection as follows: First, to address the inherent variability in the actual values of NFPs at runtime, we treat them as probability distributions. Then, on top of that, we tackle the variability needed in describing NFPs, by providing conditional contracts. Finally, from usage patterns, we compute user-specific expectations for such NFPs. Further, we depict a typical scenario, which serves both as a motivation for our approach, and as a basis for its evaluation.