Preserving performance of service computing runtime through self-tuning: A feedback control approach

Configuring and tuning modern service computing runtime becomes exceedingly complex because of the complexity of the service computing runtime architecture, the dynamic nature of service workloads and the interactions between various QoS mechanisms. In this paper, we describe a feedback control approach to make service computing runtime self-tuning. By self-tuning, we mean mechanisms to automate the tasks of configuring and tuning the runtime system so as to preserve the performance-optimize the processing delay and throughout in the system. In our approach, a feedback-driven adaptive control framework is proposed to continuously monitor the system state and adapt the system configuration based on the observed data. Based on a pipeline-based service computing runtime, we implement our framework as an adaptive control module. Experimental results demonstrate that the runtime system achieves robust performance and guarantees system stability in the presence of workload unpredictability after using our adaptive control module.