A user-focused evaluation of web prefetching algorithms Josep Dome `nech * , Ana Pont, Julio Sahuquillo, Jose ´ A. Gil Department of Computer Engineering, Universitat Polite `cnica de Vale `ncia, Camı ´ de Vera, s/n, 46022 Vale `ncia, Spain Received 14 August 2006; received in revised form 3 May 2007; accepted 11 May 2007 Available online 18 May 2007 Abstract Web prefetching mechanisms have been proposed to benefit web users by hiding the download latencies. Nevertheless, to the knowl- edge of the authors, there is no attempt to compare different prefetching techniques that consider the latency perceived by the user as the key metric. The lack of performance comparison studies from the user’s perspective has been mainly due to the difficulty to accurately reproduce the large amount of factors that take part in the prefetching process, ranging from the environment conditions to the work- load. This paper is aimed at reducing this gap by using a cost-benefit analysis methodology to fairly compare prefetching algorithms from the user’s point of view. This methodology has been used to configure and compare five of the most used algorithms in the literature under current and old workloads. In this paper, we analyze the perceived latency versus the traffic increase (both in bytes and in objects) to evaluate the benefits from the user’s perspective. In addition, we also analyze the performance results from the prediction point of view to provide insights on the observed performance. Results show that higher algorithm complexity does not improve performance, object- based algorithms outperform those based on pages, and performance among object-based algorithms present minor differences in the object traffic increase. Ó 2007 Elsevier B.V. All rights reserved. Keywords: Web prefetching; Performance evaluation; User-perceived latency 1. Introduction The knowledge and comprehension of the behavior of a web user are important keys in a wide range of fields related to the web architecture, design, and engineering. The infor- mation that can be extracted from web user’s behavior per- mits to infer and predict future accesses. This information can be used, for instance, for improving Web usability [1], developing on-line marketing techniques [2] or reducing user-perceived latency [3], which is the main goal of pre- fetching techniques. These techniques use access predictors to process a user request before the user actually makes it. Several ways of prefetching user’s requests have been proposed in the literature: the preprocessing of a request by the server [4], the transference of the object requested in advance [5,6], and the pre-establishment of connections that are predicted to be made [7]. Despite the large amount of research works focusing on this topic, comparative and evaluation studies from the user’s point of view are rare. This fact leads to the inability to quantify in a real working environment which proposal is better for the user. On the one hand, the underlying baseline system where prefetching is applied differs widely among the studies. On the other hand, different performance key metrics were used to eval- uate their benefits [8]. In addition, the used workloads are in most cases rather old, which significantly affects the pre- diction performance [9], making the conclusions not valid for current workloads. Research works usually compare the proposed prefetch- ing system with a non-prefetching one [10,3,5]. In these studies, different workload and user characteristics have been used, so making it impossible to compare the good- ness and benefits among proposals. Some papers have been published comparing the perfor- mance of prefetching algorithms [11–16]. These studies 0140-3664/$ - see front matter Ó 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.comcom.2007.05.003 * Corresponding author. Tel.: +34 963879215; fax: +34 963877579. E-mail addresses: jodode@doctor.upv.es (J. Dome `nech), apont@ disca.upv.es (A. Pont), jsahuqui@disca.upv.es (J. Sahuquillo), jagil@ disca.upv.es (J.A. Gil). www.elsevier.com/locate/comcom Computer Communications 30 (2007) 2213–2224