Raft Distributed System for Multi-access Edge Computing Sharing Resources
Zain Khaliq, Ahmed Refaey Hussein
TL;DR
The paper addresses latency and reliability challenges in sharing MEC resources across edge clusters. It combines a Raft-based consensus protocol with a private Hyperledger Fabric blockchain to coordinate task offloading and ensure data integrity, while deploying a Deep Deterministic Policy Gradient (DDPG) learner to optimize leader selection and offloading decisions. The core contributions are the integrated Raft–blockchain MEC framework and a DDPG-based training regime that yields higher rewards and lower variance in policy performance, indicating robust latency improvements. The work demonstrates that end-to-end safety, coordination, and responsiveness can be enhanced in edge computing environments through learned policies that drive resource scheduling within a blockchain-backed Raft cluster. The practical impact lies in enabling safer and more efficient MEC deployments with data integrity guarantees and adaptive latency optimization.
Abstract
Researchers all over the world are employing a variety of analysis approaches in attempt to provide a safer and faster solution for sharing resources via a Multi-access Edge Computing system. Multi-access Edge Computing (MEC) is a job-sharing method within the edge server network whose main aim is to maximize the pace of the computing process, resulting in a more powerful and enhanced user experience. Although there are many other options when it comes to determining the fastest method for computing processes, our paper introduces a rather more extensive change to the system model to assure no data loss and/or task failure due to any scrutiny in the edge node cluster. RAFT, a powerful consensus algorithm, can be used to introduce an auction theory approach in our system, which enables the edge device to make the best decision possible regarding how to respond to a request from the client. Through the use of the RAFT consensus, blockchain may be used to improve the safety, security, and efficiency of applications by deploying it on trustful edge base stations. In addition to discussing the best-distributed system approach for our (MEC) system, a Deep Deterministic Policy Gradient (DDPG) algorithm is also presented in order to reduce overall system latency. Assumed in our proposal is the existence of a cluster of N Edge nodes, each containing a series of tasks that require execution. A DDPG algorithm is implemented in this cluster so that an auction can be held within the cluster of edge nodes to decide which edge node is best suited for performing the task provided by the client.