Model and Machine Learning based Caching and Routing Algorithms for Cache-enabled Networks

TL;DR

This paper compares model-based and machine learning approaches for caching and routing in networked systems, analyzing their performance and key influencing factors through experiments and theoretical insights.

Contribution

It introduces a comprehensive comparison of model-based and machine learning methods for cache and routing strategies, highlighting their advantages and key factors affecting performance.

Findings

Content popularity skewness and request correlation significantly impact cache performance.

Routing performance is influenced by alternate path routing and content search strategies.

Machine learning models like reinforcement and deep learning are applicable for caching and routing optimization.

Abstract

In-network caching is likely to become an integral part of various networked systems (e.g., 5G networks, LPWAN and IoT systems) in the near future. In this paper, we compare and contrast model-based and machine learning approaches for designing caching and routing strategies to improve cache network performance (e.g., delay, hit rate). We first outline the key principles used in the design of model-based strategies and discuss the analytical results and bounds obtained for these approaches. By conducting experiments on real-world traces and networks, we identify the interplay between content popularity skewness and request stream correlation as an important factor affecting cache performance. With respect to routing, we show that the main factors impacting performance are alternate path routing and content search. We then discuss the applicability of multiple machine learning models, specifically reinforcement learning, deep learning, transfer learning and probabilistic graphical models for the caching and routing problem.