Performance Evaluation of Caching Policies in NDN - an ICN Architecture

TL;DR

This paper evaluates various caching policies in Named Data Networking, demonstrating that Universal Caching (UC) outperforms FIFO and LRU in different network models using simulation.

Contribution

It provides a comparative performance analysis of caching policies in NDN, highlighting UC's superiority across diverse network topologies.

Findings

UC outperforms FIFO and LRU in simulations

Universal Caching is suitable for both sensor and ISP networks

Performance varies with network topology and caching policy

Abstract

Information Centric Networking (ICN) advocates the philosophy of accessing the content independent of its location. Owing to this location independence in ICN, the routers en-route can be enabled to cache the content to serve the future requests for the same content locally. Several ICN architectures have been proposed in the literature along with various caching algorithms for caching and cache replacement at the routers en-route. The aim of this paper is to critically evaluate various caching policies using Named Data Networking (NDN), an ICN architecture proposed in literature. We have presented the performance comparison of different caching policies naming First In First Out (FIFO), Least Recently Used (LRU), and Universal Caching (UC) in two network models; Watts-Strogatz (WS) model (suitable for dense short link networks such as sensor networks) and Sprint topology (better suited for large Internet Service Provider (ISP) networks) using ndnSIM, an ns3 based discrete event simulator for NDN architecture. Our results indicate that UC outperforms other caching policies such as LRU and FIFO and makes UC a better alternative for both sensor networks and ISP networks.