Symbolic performance adaptation

Quality-of-Service attributes such as performance and reliability heavily depend on the run-time conditions under which software is executed (e.g., workload fluctuation and resources availability). Therefore, it is important to design systems able to adapt their setting and behavior due to these run-time variabilities. In this paper we propose a novel approach based on queuing networks as the quantitative model to represent system configurations. To find a model that fits with continuous changes in run-time conditions we rely on an innovative combination of symbolic analysis and satisfiability modulo theory (SMT). Through symbolic analysis we represent all possible system configurations as a set of nonlinear real constraints. By formulating an SMT problem we are able to devise feasible system configurations at a small computational cost. We study the effectiveness and scalability of our approach on a th