Genetic algorithms: an approach to optimal web cache replacement

Growing demands for increased bandwidth as a result of a media-savvy World-Wide Web (WWW) assert a need for faster and better solutions to network infrastructure. Caching objects on the web represents a software-based approach that can dramatically reduce bandwidth requirements by exploiting the repetitious nature of object requests on the web. However, limitations of disk capacity and memory constrain the size of a cache. A cache replacement algorithm allocates space for new requests by evicting objects from the cache. LRU and LRU-related strategies are known to produce the best hit ratios, given the temporal locality of URL requests on the WWW. However, LRU is not perfect and an adaptive algorithm could prove especially useful. This paper presents a novel approach to web caching by suggesting a cache replacement policy model based on an approach which utilizes genetic algorithms (GAs). The objective of the model is to produce an object of maximum cost, which is provided as input to the distance measure of each object in the cache to the object of maximum cost.