Resilient and Fast Block Transmission System for Scalable Hyperledger Fabric Blockchain in Multi-Cloud Environments

Abstract

In this paper, we propose a resilient and fast block transmission system for Hyperledger Fabric in multi-cloud environments. The goal of the proposed system is to improve the scalability, transaction throughput, and resilience of Hyperledger Fabric by minimizing the block synchronization time among nodes. To achieve this goal, the proposed system is designed to deliver blocks quickly and reliably to all the participating nodes in time-varying multi-cloud environments. The proposed system includes the delay estimating process with O(N) control message overhead over the P2P network, the effective bandwidth estimating process for block transmission, the Gaussian Mixture Model-based clustering and cluster leader selecting process, and hybrid P2P multicast tree constructing process. In addition, a control message format and delivery process are proposed to efficiently provide hybrid P2P multicast tree and neighbor nodes information to all the participating nodes. And we propose a pull-based local block loss recovery process that can receive lost blocks from multi-node without complicated scheduling using a rateless code. The proposed system is fully implemented by using well-known open sources (e.g., Hyperledger Fabric, Docker, Containernet, and Mininet) and Go/C/Python. Experiment results show that the proposed system can reduce the maximum block arriving time among all the participating nodes by approximately 50%~95% compared to the existing algorithms. This improves not only blockchain transaction per second, but also resilience to various network-layer vulnerabilities and attacks that may occur when the block propagation delay increases.