qos-aware resource allocation in distributes systems

In recent years, new computing paradigms have changed the way enterprises develop and provide their services over the Internet. In this context, Service Oriented Architecture and Cloud computing are playing a key role in leading the innovation. While the former redefines the software development process, the latter upsets the way enterprises manage the hardware and software infrastructure needed to run their services. As a consequence, new strategies are required by service providers to be able to provide these services with an adeguate Quality of Service (QoS) level. This is a critical task, especially in the context of services available over the Internet, in which QoS plays a crucial role in attracting and retaining users. In this thesis we address the problem of provide services with QoS. Our main contribution is to provide a formal solution based on optimization theory. We consider several services commonly used nowadays, namely SOA applications, Web services and computational resources provided as Virtual Machines. These services are strictly correlated each other and are representative of a wide set of common scenarios in distributed systems, which require different optimization tools. Scenarios in which a single agent in isolation optimizes its own performance, that we address using a Mixed Integer Linear Programming problem; non-cooperative scenarios in which agents selfishly compete each other to optimize their performances, that we address using a Game Theoretic approach; cooperative scenarios in which agents collaborate to maximize system efficiency, that we address using Markov Decision Process, heuristic and Reinforcement Learning.