A Parameterized Hybrid Size-LFU Replacement Algorthm for Web Caches

Abstract

Web caches serve two important purposes: to improve user-perceived response and to reduce network traffic. How well these goals are achieved depends on the caching algorithm employed and cache replacement scheme used. Substantial efforts have been made to devise suitable methods. Most existing replacement algorithms consider recency of object preferences, frequency of object references, or the fetching cost of objects. We propose a parameterized approach that combines reference pattern and object size in the design. Properties of the algorithm are analyzed. The algorithm is applied to a variety of workloads and its performance is compared against several well known methods. Preliminary results indicate that the algorithm us a viable choice.