Cooperative Content Caching for Mobile Edge Computing With Network Coding

Weisheng He, Yuhan Su, Xueting Xu, Zhaohui Luo, Lianfen Huang, Xiaojiang DuIEEE Access 2019 · DOI

Abstract

Mobile Edge Computing (MEC) is pivotal for reducing latency in high-density 5G environments by caching popular content closer to end-users. However, edge servers face significant constraints in storage capacity and update bandwidth. This paper proposes a dual-strategy optimization framework: an intelligent "compound" recommendation system for precise content selection and a novel network-coding-based relay scheme that enables cache updates to "piggyback" on standard traffic. Experimental results demonstrate a 10% reduction in cache miss rates and a significant boost in average data access speeds without requiring additional time slots for transmission.

Introduction

The proliferation of data-intensive applications like high-definition video streaming has placed immense pressure on core networks. Mobile Edge Computing (MEC) addresses this by moving content to the network edge. Despite its benefits, MEC faces two primary challenges: the "What" problem—identifying localized content preferences for limited edge storage—and the "How" problem—efficiently updating the cache without consuming dedicated bandwidth.

System Architecture

The proposed solution integrates an Intelligent Cache Analysis Platform with a Cooperative Network Coding and Relay mechanism. The architecture utilizes a Relay Station (RS-MEC) to bridge the gap between the Base Station (BS) and end-users, facilitating both normal communication and background cache updates.

System architecture of the proposed MEC caching and delivery scheme.

Intelligent Cache Analysis Platform

To optimize the selection of files for edge storage, the researchers developed a multi-stage pipeline designed to capture localized user preferences more accurately than traditional global algorithms.

Workflow of the MEC Server Intelligent Cache Analysis Platform.

The platform consists of several key components:

  • Compound Recommendation Algorithm: This algorithm moves beyond simple collaborative filtering by using a multivariate similarity formula. It accounts for numerical scores, interest tendencies, and "overlapping confidence" (the depth of shared history between users).
  • Hybrid Weighting System: Content importance is determined by combining the Entropy Method for objective data distribution analysis with the Analytic Hierarchy Process (AHP) for subjective priority logic.
  • Improved Genetic Algorithm: To solve the optimization problem of fitting high-value files into fixed storage, a genetic algorithm with a "large variation" effect is employed to avoid local optima.

Cooperative Network Coding and Relay

The delivery mechanism utilizes XOR Network Coding at the Relay Station (RS-MEC). When the Base Station transmits a standard data packet ($A$) to a user, it simultaneously sends an XOR-encoded packet ($A \oplus B$), where $B$ represents the cache update file.

The RS-MEC receives both $A$ and $A \oplus B$. By performing the operation $(A \oplus B) \oplus A$, it extracts file $B$ with perfect fidelity. Since the encoded packet is the same size as a standard packet, the cache update is essentially transmitted "for free" in terms of time slots, avoiding the need for dedicated bandwidth or idle-time windows.

Experimental Results

The system was evaluated using a custom dataset of 100 users and 50 movies, with MEC storage capped at 3GB (roughly 1/3 of the total dataset size).

  • Accuracy: The compound recommendation algorithm achieved a score deviation of only 0.29, significantly lower than user-based (0.33) or item-based (0.65) methods.
  • Efficiency: The system reduced the cache miss rate by approximately 10% compared to contrast methods.
  • Performance: In integrated simulations, the average user access rate improved from ~3.0 MB/s to 3.44 MB/s.
  • Bandwidth Savings: The network coding scheme successfully delivered cache files without increasing the number of transmission slots required for normal communication.

Conclusion

By combining human-centric recommendation logic with efficient network coding, this research provides a viable path for optimizing MEC in 5G networks. Future work will focus on scaling the Genetic Algorithm for larger libraries and exploring full-duplex capabilities to further enhance throughput.

Summary updated

BibTeX
@article{he2019cooperative,
  author = {Weisheng He and Yuhan Su and Xueting Xu and Zhaohui Luo and Lianfen Huang and Xiaojiang Du},
  title = {{Cooperative Content Caching for Mobile Edge Computing With Network Coding}},
  journal = {IEEE Access},
  year = {2019},
  doi = {10.1109/access.2019.2917977},
}
Figure