Online adaptive optimization for event-driven dynamic service composition
Baoqun Yin · Control theory & applications · 2011
An event-driven dynamic service composition strategy is presented for networked new media service systems to improve the resource utilization while simultaneously ensuring the quality of service(QoS) for multi-class services.An analytical model of semi-Markov switching state-space control processes is introduced to formulate the dynamic service composition problem.By utilizing the dynamic hierarchy of this model and the features of event-driven policy,an online optimization algorithm that combines stochastic approximation with the policy iteration is proposed.This algorithm is in-dependent of any prior knowledge of system parameters and is with less computational cost.Simulation results demonstrate the effectiveness of the proposed algorithm.