Performance evaluation for new web caching strategies combining LRU with score based object selection

The topic of Internet content caching regained relevance over the last years due to the increasing and widespread use of content delivery infrastructures to meet capacity and delay demands of multimedia services. In this study, we evaluate the performance of web caching strategies in terms of the achievable hit rate for realistic scenarios of large user populations. We focus on a class of score gated least recently used (SG-LRU) strategies which combine the simple update effort of the LRU policy with the flexibility to keep the most important content in the cache according to an arbitrarily predefined score function. Caching efficiency is evaluated via simulations assuming Zipf distributed requests, which have been confirmed manifold in access pattern of popular web platforms for video streaming and various other content types. In this paper, we analyze the hit rate gain of alternative web caching strategies for the standard independent request model (IRM) within the complete relevant range of three basic system parameters. The results confirm absolute hit rate gains of 10%–20% over pure LRU as realistic estimation in general, as experienced in other trace-driven case studies for special caching strategies. Moreover, we compare the performance for the independent request model with results for correlated dynamic request pattern over time, based on Wikipedia statistics that are available as daily top-1000 page requests. In this way, we show that the IRM analysis is valid for caches with a large user population, although high dynamics tends to reduce the achievable hit rate below the IRM result for smaller user communities.