Inference of performance annotations in web service composition models

High-quality services must keep working reliably and efficiently, and service compositions are no exception. As they integrate several internal and external services over the network, they need to be carefully designed to meet their performance requirements. Current approaches assist developers in estimating whether the selected services can fulfill those requirements. However, they do not help developers define requirements for services lacking performance constraints and historical data. Manually estimating these constraints is a time-consuming process which might underestimate or overestimate the required performance, incurring in additional costs. This work presents the first version of two algorithms which infer the missing performance constraints from a service composition model. The algorithms are designed to spread equally the load across the composition according to the probability and frequency each service is invoked, and to check the consistency of the performance constraints of each service with those of the composition.