Building design-time and run-time knowledge for QoS-based component assembly

Modern software systems are required to dynamically adapt to changing workloads, scenarios, and objectives and to achieve a certain Quality of Service (QoS). Guaranteeing QoS requirements is not trivial, as run-time uncertainty might invalidate the design-time rationale, where software components have been selected by means of off-line analysis. In this work, we propose a QoS-based feedback approach that makes a combined use of design-time predictions and run-time measurements to manage QoS data over time and support software architects while selecting software components that best fit QoS requirements. We illustrate the feasibility and efficacy of the approach on a case study, where the quantitative evaluation shows how the analysis effectively identifies the sources of QoS violations and indicates possible solutions to achieve QoS requirements.