Building Models of Computation of Service-Oriented Software via Monitoring Performance Indicators (Short Paper)

The typical pricing policy, offered by service providers, is on a pay-as-you-go basis, which is proportional to the resources consumed by services. However, providers do not expose the specification of the resource consumption of services. Thus, providers offer cost pre-calculators that are service-agnostic. Such cost pre-calculators take into account upper bounds of provided facilities, instead of an estimation of the resource consumption of a service. To help service providers in specifying the resource consumption of services, we propose an initial version of an automated approach that captures the resource consumption of services. To make it in an independent way of the used measurement units and of the infrastructure over which a service is hosted, our approach builds for a service a mathematical model of computation, estimating its computational performance that is directly related to its resource consumption. Our approach moves beyond the state-of-the-art, since it automatically builds a model of computation based on the interface of a service, instead of its source code, and based on the monitoring of the execution of a service by using its executable files, instead of the monitoring or the theoretical analysis of its source code. The evaluation of our approach in a set of basic case studies shows promising results.