Skyline-Based Multi-Criteria Cache Replacement Policy for Enhanced Performance in Named Data Networking

Abstract

Named Data Networking (NDN) represents a promising future Internet architecture that addresses the limitations of current IP-based networks through distributed content caching. Cache replacement policies play a crucial role in determining the overall performance of NDN networks. While existing policies such as LFU, LRU, and recent hybrid approaches like LPCE have shown improvements, they often optimize for a single metric and may not achieve optimal performance across multiple criteria simultaneously. This paper introduces SkyCache, a novel cache replacement scheme based on the Skyline operator from database systems, which enables multi-criteria optimization for content eviction decisions in NDN routers. Our approach considers multiple dimensions including popularity, recency, size, and retrieval cost to identify non-dominated content objects for caching. Through extensive simulations using the ccnSim simulator across various network topologies (TREE and CDN), we demonstrate that SkyCache outperforms existing policies including the recent LPCE approach. Experimental results show that SkyCache achieves a cache hit ratio improvement of 2.5% to 7.8% compared to LPCE, with corresponding reductions in content delivery delay (8-15%), network traffic (12-18%), and producer load (10-16%).