Performance Models for Caching World Wide Web Traffic

摘要

The performance of a cache can be studied by either a simulation using an actualtrace of traffic, simulation using a trace based on a model or with the model itself. The performance of a cache depends on the users' requests, cache size and replacement algorithm. This thesis examines three stochastic arrival models, Independent Reference Model (IRM), Least Recently Used (LRU) stack model and renewal model. Simulations of renewal model suggest that it is not optimal for finding ways to improve cache performance. Results of trace simulations of suggested replacement algorithms are reviewed and found hard to compare with each other.