Feedback Control-Based QoS Guarantees in Web Application Servers

Abstract

This paper considers providing two types of QoS guarantees, proportional delay differentiation (PDD) and absolute delay guarantee (ADG), in the database connection pool (DBCP) for Web application servers using the classical feedback control theory. PDD aims to maintain the average queuing delay ratio between different classes of requests according to pre-specified parameters, and ADG aims to ensure the average queueing delay for requests with high priority is no more than the threshold configured. To achieve these goals, we establish the approximate linear time-invariant models of the DBCP through system identification experimentally, and design two proportional-integral (PI) controllers, PDD controller and ADG controller, using the root locus method. These controllers are invoked periodically to calculate and adjust the probabilities for different classes of requests to use a limited number of database connections, according to the error between the measured QoS metric and the reference value. We implement all components of the closed-loops in a real DBCP for Web application servers, and design three kinds of workloads, which follow deterministic, uniform and heavy-tailed distributions respectively, to evaluate the performance of the closed-loop systems. Experiment results demonstrate that, the controllers designed are effective in handling varying workloads, PDD and ADG can be achieved in the Web application server even if the number of concurrent requests changes abruptly.