Skip to Main Content
A widely existing problem in contemporary web application servers is the delay of response time. A long response time will cause the loss of clients and the drop of commercial interest for web site operators. The web Quality of Service (QoS) control should be deployed in web application servers in order to meet different requirements from clients and applications. In this paper, we use classical feedback control theory to contribute a QoS guarantee, i.e. the absolute delay guarantee (ADG). ADG is designed to ensure that the average queuing delay for the requests with high priority is no more than the threshold configured. The approximate linear time-invariant model of the database connection pool (DBCP) in Tomcat web application server is implemented through system identification experimentally. ADG controller is designed to use Root Lacus method. According to the error between the measured queuing delay and the reference value, the controller is periodically invoked to calculate and adjust the probability for different classes of requests to use a limited number of database connections. All components of the closed-loops of QoS guarantee is implemented for HTTP dynamic requests in the database connection pool. In order to evaluate the performance of the closed-loop systems, we design some experiments and gain experiment results. Experiment results show that the controller we design are able to handle fluctuate workloads effectively. ADG can be achieved well in the DBCP in Tomcat Web Server even if the number of concurrent requests changes abruptly.