Computation Offloading Based on a Distributed Overlay Network Cache-Sharing Mechanism in Multi-Access Edge Computing

Yazhi Liu, Zhong Peng-fei, Zhigang Yang, Wei Li, Siwei LiFuture 2024 · DOI

TL;DR

This paper introduces the Distributed Edge Service Caching and Offloading (DESCO) architecture, which uses a novel "Cache Chord" mechanism for decentralized cache sharing among edge servers. By modeling computation offloading as a game theory problem and solving it with a multi-dimensional particle swarm optimization algorithm, they achieve over 27% reduction in user energy consumption while meeting real-time latency demands.

What Problem Does This Paper Address?

Modern applications like connected vehicles, AR/VR, and smart cities demand low latency and high computational power, which traditional cloud computing struggles to provide due to network latency and fluctuating quality. Multi-Access Edge Computing (MEC) aims to solve this by bringing computation closer to users. However, existing MEC solutions face challenges with limited edge device storage, communication bandwidth, and wireless coverage, hindering efficient cache reuse and energy savings. Specifically, coordinating computation offloading and cache management across distributed edge servers to minimize user energy consumption without violating strict real-time latency constraints is a significant hurdle.

Methods

The authors propose a novel Distributed Edge Service Caching and Offloading (DESCO) architecture. Key components include:

  • Cache Chord: A decentralized resource-sharing mechanism based on consistent hashing. It organizes edge servers into a circular logical topology, enabling efficient data sharing and retrieval across geographically distributed nodes. This is further enhanced by a cache IP list for mapping application resources to physical storage locations.
  • Real-Time Computation Offloading (RCO) Problem Formulation: The RCO problem is framed as a multi-player static game. This approach acknowledges the competitive nature of users selecting offloading channels, where choices impact both individual benefits and overall system energy consumption.
  • Nash Equilibrium (NE) Analysis: The paper proves the existence of Nash equilibrium solutions for the formulated game, ensuring that a stable strategy exists where no player can unilaterally improve their outcome.
  • Multi-Dimensional Discrete Particle Swarm Optimization (MDPSO): To find the optimal RCO strategy and achieve the Nash equilibrium, a multi-dimensional discrete PSO algorithm is employed. This algorithm explores the solution space to identify the offloading strategy that minimizes user energy consumption while satisfying latency constraints.

Datasets

Not discussed in the paper. The authors mention simulation experiments were conducted using a Python platform and a simulated distributed environment.

Key Insights and Ideas

  • Decentralized Cache Sharing is Key: The "Cache Chord" mechanism demonstrates that a structured, decentralized approach to cache sharing among edge servers can significantly improve resource utilization and reduce the need for users to upload raw data.
  • Game Theory for Offloading Decisions: Framing computation offloading as a multi-player static game provides a robust theoretical foundation for understanding user interactions and finding optimal, stable offloading strategies.
  • MDPSO for Complex Optimization: The MDPSO algorithm proves effective in navigating the complex, multi-dimensional search space of offloading decisions, balancing energy consumption and latency constraints.
  • Collaboration Outperforms Isolation: The results highlight that collaborative MEC server networks (using Cache Chord) offer superior performance compared to isolated servers, especially as the number of users and tasks increases.

Key Results

  • The proposed DESCO architecture with the Cache Chord mechanism and MDPSO algorithm achieved over 27% reduction in average user energy consumption compared to existing algorithms.
  • The MDPSO algorithm showed better convergence than the Greedy strategy, with an average reduction of 19.7% in user-level energy consumption per round in single-server scenarios.
  • In simulations, the MDPSO algorithm achieved an average latency of 7.4 ms, significantly outperforming the Greedy strategy (11.4 ms) and local computing.
  • Increasing the number of MEC servers in the Cache Chord network led to a significant improvement in the cache hit ratio, reaching over 91% with 10 servers.
  • The MDPSO-based multi-server collaboration reduced average energy consumption by 7.89% to 4.7% compared to single-node modes.

Limitations and Future Work

  • Backhaul Link Latency: The paper acknowledges that as the number of server nodes increases, the transmission delay of corresponding backhaul links will also increase, posing new challenges for performance optimization.
  • Real-World Deployment: The study was conducted through simulations. The authors suggest the next step will be to explore how larger node scales affect service quality and to deploy the protocol in a real network environment to test its performance.
  • Parameter Sensitivity: While Zipf distribution was used for user requests, the impact of other request patterns or varying network conditions (e.g., dynamic bandwidth) could be further explored.

Summary updated

Abstract

Multi-access edge computing (MEC) enhances service quality for users and reduces computational overhead by migrating workloads and application data to the network edge. However, current solutions for task offloading and cache replacement in edge scenarios are constrained by factors such as communication bandwidth, wireless network coverage, and limited storage capacity of edge devices, making it challenging to achieve high cache reuse and lower system energy consumption. To address these issues, a framework leveraging cooperative edge servers deployed in wireless access networks across different geographical regions is designed. Specifically, we propose the Distributed Edge Service Caching and Offloading (DESCO) network architecture and design a decentralized resource-sharing algorithm based on consistent hashing, named Cache Chord. Subsequently, based on DESCO and aiming to minimize overall user energy consumption while maintaining user latency constraints, we introduce the real-time computation offloading (RCO) problem and transform RCO into a multi-player static game, prove the existence of Nash equilibrium solutions, and solve it using a multi-dimensional particle swarm optimization algorithm. Finally, simulation results demonstrate that the proposed solution reduces the average energy consumption by over 27% in the DESCO network compared to existing algorithms.
BibTeX
@inproceedings{liu2024computation,
  author = {Yazhi Liu and Zhong Peng-fei and Zhigang Yang and Wei Li and Siwei Li},
  title = {{Computation Offloading Based on a Distributed Overlay Network Cache-Sharing Mechanism in Multi-Access Edge Computing}},
  booktitle = {Future},
  year = {2024},
  doi = {10.3390/fi16040136},
}
Figure