Predicting Quality of Service of Software Applications

The ability to a priori predict the Quality of Service (QoS) of a software application is crucial both in the design of applications and in the definition of their Service Level Agreements (SLA). QoS prediction is challenging because of the different possible results of service invocations, and of the nondeterminism, correlations and complex dependencies among activities. In this thesis we present a technique to probabilistically predict the QoS of service based and parallel design patterns based applications by applying Monte Carlo simulations to a simple representation of the control-flow of the applications. A proof-of-concept implementation of the analyses is presented and applied to various examples.