IEEE TRANSACTIONS ON BROADCASTING, VOL. 54, NO. 3, SEPTEMBER 2008 599 Revenue Optimized IPTV Admission Control Using Empirical Effective Bandwidth Estimation Alan Davy, Member, IEEE, Dmitri Botvich, Member, IEEE, and Brendan Jennings, Member, IEEE Abstract—The paper presents an admission approach for IPTV service providers that is designed to minimize QoS violations whilst effectively utilizing available bandwidth. Central to the approach is an empirical method of estimating the effective bandwidth required to satisfy QoS targets for admitted traffic flows. The paper describes this method and specifies two admis- sion control algorithms based on the use of effective bandwidth estimates. The first algorithm employs a simple evaluation of whether there is sufficient bandwidth available to ensure, with an appropriate degree of confidence, that QoS targets will not be violated if a requested flow is admitted. The second algorithm utilizes information relating to the cost, duration and request fre- quency of specific IPTV content to prioritize higher revenue flows within the admission control process. Results of a simulation study (employing real traffic traces of long-lived flows) indicate that the proposed algorithms ensure that an adequate, but not overly generous, amount of bandwidth is allocated to ensure that QoS targets for accepted flows are met. Furthermore, they demonstrate the potential advantage of using content specific information in the admission control process to maximize generated revenue. Index Terms—Admission control, aggregated traffic, effective bandwidth, quality of service. I. INTRODUCTION A DMISSION control is a technique used by service providers to ensure customers’ traffic flows are allocated sufficient bandwidth to ensure service level agreement con- straints, relating to packet-level Quality-of-Service (QoS), are maintained during periods of high network load. The goal of the service provider is to ensure QoS for accepted traffic flows whilst maximizing bandwidth available for newly arriving flows. For Internet Protocol Television (IPTV) service delivery traffic flows are generally streaming movies or TV programs, which are typically high-bandwidth and have associated with them stringent QoS targets. Therefore, admission control plays a vital role, since bad admission decisions can significantly degrade QoS, not only for the newly accepted flow, but also for already accepted and ongoing flows. A key feature of any admission control algorithm is how well it predicts the level of resources required to admit a requesting Manuscript received September 30, 2008; revised April 23, 2008. Published August 20, 2008 (projected). This work is supported in part by the Science Foundation Ireland, via the Autonomic Management of Communications Net- works and Services program under Grant 04/IN3/I4040C) and also in part by the 2005 Research Frontiers project Accounting for Dynamically Composed Services under Grant CMS006). The authors are with the Telecommunications Software & Systems Group, Waterford Institute of Technology, Waterford, Ireland (e-mail: adavy@tssg.org; dbotvich@tssg.org; bjennings@tssg.org). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TBC.2008.2001245 flow. If a new flow request is accepted, the associated packet level QoS targets should be met, without affecting QoS of al- ready admitted flows. Fundamental to achieving this is the ac- curate prediction of required effective bandwidth (the minimum amount of bandwidth required by a traffic stream to maintain specified QoS related targets [1]) of the aggregated traffic al- ready admitted to the network. If the admission control algo- rithm predicts that aggregated effective bandwidth should the flow be accepted will be greater than the available bandwidth, the flow should be rejected. Clearly, the more accurate the es- timation of effective bandwidth, the more effectively the ad- mission control algorithm operates. As described later, current effective bandwidth estimation approaches are typically based on theoretical analyses of traffic properties (see for example [1]–[3]) and make simplifying assumptions such as constant packet sizes and inter-arrival times. To overcome the limitations of existing effective bandwidth estimation approaches we have proposed an empirical approach, which we use in this paper as the basis of two IPTV-focused admission control algorithms. The first, which we term Empir- ical Admission Control (EAC), provides an accurate means of predicting the amount of bandwidth required to ensure admitted traffic flows will maintain agreed packet-level QoS targets for the consequent aggregated set of traffic flows. The second, which we term Revenue Maximizing Empirical Admission Control (RMEAC), extends EAC by using information relating to the cost, duration and request arrival rate of IPTV content to provide an admission control regime that seeks to maximize the revenue that is generated for the service provider by the accepted flows. We compare the performance of EAC with respect to bandwidth utilization and QoS control to a number of comparative admission control (AC) algorithms, through simulation of a simple IPTV network focusing on a service provider wishing to control QoS over limited bandwidth. The paper is organized as follows: Section II provides background information on the following areas: IPTV service delivery; the effect QoS violations has on IPTV services and customer perception of quality; alternative approaches to con- trolling QoS for IPTV services; admission control techniques in general; and approaches to estimating effective bandwidth. Section III defines our admission control framework, outlining the effective bandwidth estimation approach and specifying the EAC and RMEAC algorithms. Section IV defines the IPTV architecture studied, discussing topology settings, traffic characteristics, user profiles, and QoS violation measure- ment. Section V details three admission control algorithms for comparison namely Parameter Based Admission Control (PBAC) [4], [5], Experience based Admission Control (EBAC) 0018-9316/$25.00 © 2008 IEEE