Conference Paper

A Class-Based Least-Recently Used Caching Algorithm for World-Wide Web Proxies

DOI: 10.1007/978-3-540-45232-4_17 Conference: Computer Performance Evaluations, Modelling Techniques and Tools. 13th International Conference, TOOLS 2003, Urbana, IL, USA, September 2-5, 2003, Proceedings
Source: DBLP


In this paper we study and analyze the influence of caching strategies on the performance of WWW proxies. We propose a new strategy, class-based LRU, that works recency- as well as size-based, with the ultimate aim to obtain a well-balanced mixture between large and small documents in the cache, and hence, good performance for both small and large object requests. To achieve this aim, the cache is partitioned in classes, each one assigned to a specific document size range; within a class, the classical LRU strategy is applied.
We show that for class-based LRU good results are obtained for both the hit rate and the byte hit rate, if the size of the classes and the corresponding document size ranges are well chosen. The latter is achieved by the use of a Bayesian decision rule and a characterisation of the requested object-size distribution. In doing so, class-based LRU is an adaptive strategy: a change in request patterns results, via a change in the distributions, in a change in cache partitioning and request classification. Finally, the complexity of class-based LRU is comparable to that of LRU and, therefore, smaller then of its “competitors”.

5 Reads
  • Source
    • "–[19]. We studied the sizes of the objects requested by the clients. "
    [Show abstract] [Hide abstract]
    ABSTRACT: A typical step in the model-based evaluation of communication systems is to fit measured data to analytically tractable distributions. Due to the increased speed of today's networks, even basic measurements, such as logging the requests at a Web server, can quickly generate large data traces with millions of entries. Employing complex fitting algorithms on such traces can take a significant amount of time. In this paper, we focus on the Expectation Maximization-based fitting of heavy- tailed distributed data to hyper-exponential distributions. We present a data aggregation algorithm which accelerates the fitting by several orders of magnitude. The employed aggregation algorithm has been derived from a sampling stratification technique and adapts dynamically to the distribution of the data. We illustrate the performance of the algorithm by applying it to empirical and artificial data traces.
    Full-text · Conference Paper · Mar 2008
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: The World Wide Web has undergone major changes in recent years. The idea to see the Web as a platform for services instead of a one-way source of information has come along with a number of new applications, such as photo and video sharing portals and Wikis. In this paper, we study how these changes affect the nature of the data distributed over the World Wide Web. To do so, we compare two data traces collected at the web proxy server of the RWTH Aachen. The first trace was recorded in 2000, the other more than seven years later in 2007. We show the major differences, and the similarities, between the two traces and compare our observations with other work. The results indicate that traditional proxy caching is no longer effective in typical university networks.
    Full-text · Conference Paper · Sep 2008
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: The Quality of Services (QoS) guaranteed by the Internet Service Providers (ISPs) is an important factor for users’ satisfaction in using the Internet. The implementation of the web proxy caching has been implemented to support this objective and also support the security procedure of the organizations. However, the success of guaranteeing the QoS of each ISP must be depended on the cache size and efficient caching policy. This paper proposes a new architecture of cache farming with the recommender system concept to manage users’ requirements. This solution helps reducing the retrieval time and also increasing the hit rate although the number of users increases without expanding the size of caches in the farm.
    Preview · Article · Jun 2009
Show more