Pricing-based strategies for autonomic control of web servers for time-varying request arrivals

Abstract

This paper considers a web service that receives requests from customers at various rates at different times. The objective is to build an autonomic system that is tuned to different settings based on the varying conditions, both internally and externally. The authors have developed revenue-based pricing as well as admission control strategies, taking into account quality of service issues such as slow down and fairness aspects. Three heuristics are developed in this paper to address the pricing and admission control problem. The three heuristics are: (1) static pricing combined with queue-length-threshold-based admission control; (2) dynamic optimal pricing with no admission control; and (3) static pricing with nonnegative-profit-based admission control. These three strategies are benchmarked against a fourth strategy (called—do nothing) with no pricing and no admission control. The paper evaluates and compares their performance, implementability and computational complexity. The conclusion is that the web server revenue can be significantly increased by appropriately turning away customers via pricing or admission control mechanisms, and this can be done autonomically in the web server.