A nonlinear feedback control approach for differentiated performance management in Autonomic systems

Management of Quality of Service (QoS) performance and resources at runtime is much researched area. In particular, when there are different classes requiring different levels of QoS guarantees raise many research challenges. The relative/proportional guarantee scheme while dynamic resource management is proposed to implement such differentiated services in software systems. The existing approaches utilize linear feedback control methodologies to achieve the performance goals of relative guarantee scheme. In this paper, the performance issues of linear feedback control due to highly nonlinear nature of that scheme are investigated. Then, a nonlinear control theoretic approach called block oriented Hammerstein model is proposed to capture the nonlinearities. In addition, the proposed approach provides systematic/formal controller design process without requiring any modification to the managed software system/component. The experiments conducted on many different conditions illustrates that the proposed nonlinear approach significantly improves the autonomic management of differentiated QoS and resources compared to linear feedback control.