Video on Demand Streaming Using RL-based Edge Caching in 5G Networks

Rasoul Nikbakht,Sarang Kahvazadeh,Josep Mangues-Bafalluy

DOI: https://doi.org/10.48550/arXiv.2211.13962

2022-11-25

Abstract:Edge caching can significantly improve the 5G networks' performance both in terms of delay and backhaul traffic. We use a reinforcement learning-based (RL-based) caching technique that can adapt to time-location-dependent popularity patterns for on-demand video contents. In a private 5G, we implement the proposed caching scheme as two virtual network functions (VNFs), edge and remote servers, and measure the cache hit ratio as a KPI. Combined with the HLS protocol, the proposed video-on-demand (VoD) streaming is a reliable and scalable service that can adapt to content popularity.

Networking and Internet Architecture

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to improve the performance of Video on Demand (VoD) services in 5G networks through edge - caching techniques based on Reinforcement Learning (RL), especially in terms of reducing latency and backhaul traffic. Specifically, the paper proposes a caching strategy that adapts to time - and location - dependent content popularity patterns, aiming to optimize the video transmission efficiency in 5G networks.

The main contributions of the paper include:
- Developing a Reinforcement - Learning - based framework for edge caching. This framework can define states according to the number of requests for cached content and its ID, and uses advanced Soft Actor - Critic Networks to obtain efficient caching strategies.
- Deploying an end - to - end 5G network, where the core part is located in the cloud and the caching servers are deployed at the edge.
- The proposed solution is scalable, cloud - native, and can be automatically deployed at the network edge using the Open Source Management and Orchestration (OSM) framework.

Through these methods, the paper shows how to use commercial devices (such as smart phones or laptops) as video - playing tools and verifies the effectiveness of Reinforcement - Learning - based edge caching in reducing latency and backhaul traffic. In addition, this system can also adapt to the popularity changes of different contents and is superior to traditional caching techniques, such as the Least Frequency Used (LFU) strategy.