ACES: An efficient admission control scheme for QoS-aware web servers Xiangping Chen, Huamin Chen, Prasant Mohapatra * 2063 Engineering II, Department of Computer Science, University of California, One Sheilds Ave., Davis, CA 95616, USA Received 22 May 2002; revised 9 October 2002; accepted 31 October 2002 Abstract The unpredictability of server response performance hinders the advance of new application on the Internet. In this paper, we present an efficient admission control algorithm, ACES, based on the server workload characteristics. The admission control algorithm ensures the bounded response time from a web server by periodical allocation of system resources according to the resource requirements of incoming tasks. By rejecting requests exceeding server capacity, the response performance of the server is well maintained even under high system utilization. The resource requirements of tasks are estimated based on their types. A double-queue structure is implemented to reduce the effects caused by estimation inaccuracy, and to exploit the spare capacity of the server, thus increasing the system throughput. The admission control algorithm can be used for server overload control and for QoS provisioning of service differentiating Internet servers. Response delays of accepted tasks are bounded by the desired predefined time period. Theoretical analyses and experimental studies show that the ACES algorithm provides desirable throughput and bounded response delay to the tasks, without any significant impact on the aggregate throughput performance of the system under various workload situations. q 2002 Elsevier Science B.V. All rights reserved. Keywords: Admission control; Bounded response time; Internet; Admission control based on estimation of service time; QoS; Service differentiating Internet servers 1. Introduction The Internet and its services, especially the use of World Wide Web (WWW) in commercial activities, well known as e-commerce, are increasing explosively [1].A widely existing problem in contemporary web servers, however, is the unpredictability of response time. Usually, one second response time is desired from web sites, which is appropriate to the human response speed [2]. Long response delay frustrates user interest in interaction with web sites, thus devalues the service quality. Although current web servers are able to serve thousands of requests per second, the response delay of a popular server can be several seconds even minutes during high load periods, causing the de facto ‘denial-of-service’ effects. It was estimated that in 1998 about 10–25% of e- commerce transactions were aborted owing to long response delay, which translated to about 1.9 billion dollars loss of revenue [3]. The unpredictability of the web response is mainly due to temporary server overload conditions. The peak workload of a web server is usually several orders of magnitude higher than the average load, and the average load usually experiences continuous growth, making it expensive and even impossible to plan the server capacity to fulfill service requirements at all times. An overloaded server does not have enough resources to serve all the incoming requests, and has to deny service to some clients. Furthermore, it wastes resources in processing non-productive operations, such as rejecting new connections or aborting partially completed ones, while failing to process tasks to com- pletion. Abdelzaher and Bhatti [4] reported that as much as half of the server’s processing capacity is wasted on eventually rejected requests when the load is three times the server capacity. Admission control (AC) has been used for overload protection. However, most contemporary web servers use a rather naive AC scheme, namely tail-dropping AC, in which incoming requests are dropped when the number of awaiting tasks exceeds a predefined threshold. The tail-dropping AC scheme requires careful system capacity planning and works well only in steady workload situations where 0140-3664/03/$ - see front matter q 2002 Elsevier Science B.V. All rights reserved. PII: S0140-3664(02)00259-1 Computer Communications 26 (2003) 1581–1593 www.elsevier.com/locate/comcom * Corresponding author. Tel.: þ 1-530-7548380; fax: þ 1-530-7524767. E-mail addresses: prasant@cs.ucdavis.edu (P. Mohapatra), chenhua@ cs.ucdavis.edu (H. Chen).