An Adaptive Neural Network-Based Method for Tile Replacement in a Web Map Cache

Most popular web map services, such as Google Maps, serve pre-generated image tiles from a server-side cache. However, storage needs are often prohibitive, forcing administrators to use partial caches containing a subset of the total tiles. When the cache runs out of space for allocating incoming requests, a cache replacement algorithm must determine which tiles should be replaced. Cache replacement algorithms are well founded and characterized for general Web documents but spatial caches comprises a set of specific characteristics that make them suitable to further research. This paper proposes a cache replacement policy based on neural networks to take intelligent replacement decisions. Neural networks are trained using supervised learning with real data-sets from public web map servers. Hight correct classification ratios have been achieved for both training data and a completely independent validation data set, which indicates good generalization of the neural network. A benchmark of the performance of this policy against several classical cache management policies is given for discussion.