Understanding the Potential Benefits of Cooperation among Proxies: Taxonomy and Analysis

Cooperation among proxies is potentially beneficial in improving the Web performance for two reasons: First, it reduces capacity misses by effectively increasing total cache size. Second, it reduces compulsory misses by opportunistically finding the missed object in another proxy’s cache. While several studies have investigated the benefit of cooperative caching, they vary significantly in their conclusions. The goal of this study is to determine the potential benefits achievable through cooperation among nearby proxies under various network configurations and user access behaviors. We begin this study by providing a taxonomy, which categorizes existing cooperative caching schemes into three broad forms based on the level of cooperation involved. We then provide a simple mathematical model to quantify the upperbound of performance gain achievable through cooperation. Finally, we use a trace-driven simulation to validate the upperbounds derived from the analysis, and show that the actual gain of cooperation is quite sensitive to the user access behavior. Our results corroborate the findings of previous studies and explain their results in a unified analytical framework. In practice, the findings in this paper can be used to suggest the proper level of cooperation that can reap the most benefit of cooperative caching given the network configuration and user access behavior.