Feedback control approaches for quality of service guarantees in Web servers

It is important to guarantee quality of service (QoS) in heavy-loaded Web servers. In this paper, we present a framework eQoS that monitors and controls the service quality delivered to clients in Web servers. Because existing approaches have weak control over delivered service quality, we present two control approaches to manage server resources: a model-based approach based on queuing theory and feedback control theory; and a model-free control approach that based on fuzzy control theory within the framework. The model-based approach is desirable when Web servers can be characterized as an M/G/1 queuing system. Otherwise, the model-free approach is appropriate. Experiments have been conducted to demonstrate the effectiveness of these two control approaches. We compared the model-free approach with queuing-theoretic approach. We also carried out experiments to compare the model-free approach with static and adaptive feedback control approaches and static fuzzy control approach. Experimental results show that the model-based and model-free approaches outperform their competitors with smaller variance.