Prediction-Based Admission Control Schemes for Proportional Differentiated Services Enabled Internet Servers

Abstract

A common problem in contemporary web servers is the unpredictability of response time. Researchers have recently considered different admission control algorithms with differentiated service for web servers to complement the Internet differentiated service model, and thereby provide QoS support to users of the World Wide Web. However, most of these admission control mechanisms do not ensure the QoS requirements of all admitted clients under a bursty workload. Although enabled in web servers to improve the QoS guarantee predicament in previous literature, an Internet service model called proportional differentiated service remains impractical and incompatible with current Internet protocols. This work presents two algorithms for admission control and traffic scheduling algorithms of the web server under proportional differentiated service, which embed a time series predictor to estimate the client traffic load in the next time period. The time series predictor is implemented using four different approaches based on their successful prediction rate in the literature. The experimental results demonstrate that the proposed approaches can effectively realize proportional delay differentiation service in multiclass Web servers when the support vector machines algorithm is utilized as the time-series predictor.