QoS-Aware Adaptive Service Orchestrations

Abstract

Service Oriented Computing enables distributed \napplications that orchestrate existing services exported by remote \nproviders. This paradigm requires to explicitly handle possible \nchanges that may affect the orchestration. They include changes \nthat impact its functional behavior (e.g., services being retired by \ntheir providers), but also changes in the non-functional behavior \nof the orchestrated services (e.g., an increased execution time). In \nthe past we developed DSOL: it combines a declarative language \nto model the orchestration with planning mechanisms to decide at \nrun-time the best flow of actions. In this paper we extend DSOL \nto support QoS attributes and requirements. In particular, we \ncombine the DSOL planning techniques with a linear optimizer to \ncalculate the optimal plan w.r.t. the QoS requirements. Moreover, \nwe leverage the DSOL ability to adapt the orchestration flow at \nrun-time, to further optimize the QoS perceived by the end users \ndepending on the actual situations encountered.