Model-based dynamic QoS-driven service composition

As a consequence of their ever increasing pervasiveness in today's systems, software services are expected to guarantee their QoS even when operating in contexts whose operational conditions may continuously change. To cope with such continuous change, services must evolve in a way that is transparent to the end-user. This can be done by exploiting sophisticated means to reason about quality of service (QoS) and to drive service construction in a dynamic and automated fashion. The paper tackles this challenging scenario by proposing a model-based framework, called smart, that automatically constructs complex services with guaranteed QoS. smart exploits rich service descriptions (in particular concerning QoS characterizations) to automate the negotiation of Service Level Agreements (SLA) and to realize SLA-driven automated service reconfiguration, when the target QoS cannot be achieved by the current service assembly. Finally, smart addresses the case in which some services involved in the composition do not support SLA negotiation. In this case, monitoring is used to characterize empirically (as opposed to ``contractually'') the QoS offered by such non-guaranteed services