In-Network Caching and Content Placement in Cooperative Small Cell Networks

Francesco Pantisano, Mehdi Bennis, Walid Saad, Mérouane Debbah

Abstract

This paper introduces a decentralized, cooperative caching framework designed for small cell base stations (SBSs) to mitigate backhaul congestion in modern cellular networks. By enabling SBSs to form coalitions and coordinate content placement, the proposed approach optimizes storage utilization and reduces reliance on the core network. Simulation results demonstrate up to a 21% improvement in quality-of-service (QoS) and content availability compared to existing uncoordinated caching methods.

Introduction

The exponential growth of mobile multimedia traffic has placed immense strain on the "backhaul"—the critical link connecting local base stations to the Evolved Packet Core (EPC). While Small Cell Base Stations (SBSs), such as femtocells and picocells, bring network resources closer to users, their limited backhaul capacity often remains a bottleneck.

Traditional caching strategies at the SBS level typically operate independently. This "selfish" approach leads to significant inefficiencies:

  1. Content Duplication: Neighboring SBSs often cache identical popular files, wasting local storage.
  2. Unbalanced Load: If a file is missing locally, the SBS must fetch it from the EPC, even if a nearby neighbor already possesses the content.

To address these issues, this research proposes a distributed "in-network" caching system where SBSs pool their storage resources to minimize core network traffic.

Proposed Cooperative Framework

The authors frame content placement as a cost-minimization problem. They define a hierarchy of retrieval costs: zero for local cache, low for neighboring SBSs within a coalition, and high for the core network.

System architecture showing cooperative in-network caching in Network Domain A versus non-cooperative caching in Network Domain B.

Instead of relying on a complex central controller, the framework employs a decentralized coalition formation process. SBSs within the same network domain identify neighbors and coordinate their caches through a two-phase heuristic algorithm:

  • Discovery Phase: SBSs identify neighbors and prioritize them based on available backhaul bandwidth.
  • Management Phase: SBSs exchange proposals to maximize "cooperative utility," which represents the total backhaul cost saved for the entire group.

Mathematical analysis confirms that this algorithm has polynomial complexity and provides a performance lower bound, ensuring computational feasibility for large-scale deployments.

Performance Evaluation

The framework was evaluated in a simulated environment featuring a 500x500m macro-cell area with up to 200 SBSs. Traffic patterns followed a Zipf distribution (skewness $\psi = 1.12$), modeling the typical skewness of file popularity.

Key Results

The cooperative approach significantly outperforms traditional methods. As shown in the results, the proposed in-network caching approach maintains a higher average utility per SBS across various backhaul capacities.

Average utility per SBS as a function of backhaul capacity, comparing the proposed cooperative approach against FemtoCacher and greedy methods.

Specifically, the proposed method achieved a 21% gain in utility over "FemtoCacher" (a state-of-the-art geographical caching scheme) and a 45% gain over standard greedy uncoordinated approaches. Furthermore, the system scales effectively; the number of iterations required for convergence remains low even as the number of SBSs increases.

Conclusion and Future Work

By prioritizing cache diversification over duplication and accounting for residual backhaul bandwidth, this cooperative framework effectively increases the total content library available at the network edge.

Future research directions include adapting the model for dynamic user mobility, as the current study assumes stationary users. Additionally, integrating real-time popularity estimation would enhance the system's robustness compared to the static Zipf distribution used in these simulations.

Summary updated

BibTeX
@inproceedings{pantisano2014network,
  author = {Francesco Pantisano and Mehdi Bennis and Walid Saad and Mérouane Debbah},
  title = {{In-Network Caching and Content Placement in Cooperative Small Cell Networks}},
  booktitle = {5GU},
  year = {2014},
  doi = {10.4108/icst.5gu.2014.258230},
}
Figure