Skip to Main Content
Computer systems are becoming more complex continuously and managing them is quite challenging even to the most skilled IT professionals. A promising way to address obstacles to efficient management and re-organization of computer systems relies on the automation of these tasks. An approach has been introduced that in which analytic performance models are combined with combinatorial search techniques to design a QoS controller that runs periodically to determine the best possible configuration for the system given its workload. The results of applying this approach have been shown to QoS control of many systems. In this paper we considered a multithreaded web server with multiple classes of request and proposed that the QoS controller re-configure it in each controller interval distinguishing different workload classes. We illustrate that this mechanism enhances the self-optimizing and self-managing capability of the system, and improves global QoS value of it using simulated multithreaded web server. In addition, we show that applying this mechanism make QoS controller to become fairer and more flexible.
Date of Conference: July 28 2008-Aug. 1 2008