Incremental Least-Recently-Used Algorithm: Good, Robust, and Predictable Performance

In this paper, we propose a replacement algorithm for file caching in communication networks. While various schemes for file replacement exist, achieving simultaneous good, robust, and predictable performance remains challenging. To address this challenge, we introduce the Incremental Least-Recently-Used (iLRU) algorithm, a general scheme building on the Least-Recently-Used (LRU) algorithm. iLRU initially caches a portion of a file upon the first request and incrementally caches more with subsequent requests. Request frequency is inferred from caching states without additional overhead, where a larger cached size indicates higher request frequency. We develop an analytical model based on the Time- to- Live (TTL) concept to predict iLRU's hit ratio. Our theoretical analysis demonstrates that iLRU allocates a larger cache space to popular files, achieving a higher hit ratio than the LRU algorithm. Simulation results show that iLRU outperforms classical algorithms and validate the accuracy of the analytical model. Furthermore, the results on real traces validate iLRU's robustness in various scenarios.