FAULT-TOLERANT DATA DISTRIBUTION IN EDGE COMPUTING VIA ERASURE CODING: THE EDGE HYDRA APPROACH

Abstract

In edge computing environments, distributing popular data from the cloud to edge servers efficiently is critical for providing low-latency access. Existing distribution schemes divide a file into multiple data blocks and transmit them in parallel to target edge servers, which then reconstruct the file upon receiving all blocks. Although this approach accelerates data delivery, it is vulnerable to transmission delays and failures caused by network fluctuations or server outages, as the successful reconstruction depends on receiving every data block. To address these limitations, this paper proposes EdgeHydra, a fault-tolerant data distribution scheme based on erasure coding. EdgeHydra encodes a file into both data and parity blocks, allowing edge servers to reconstruct the original file from any sufficient subset of these blocks without waiting for all transmissions to complete. Furthermore, it employs a leaderless block supplement mechanism, enabling distributed coordination among edge servers to recover missing blocks without relying on a centralized controller. Experimental results demonstrate that EdgeHydra significantly improves robustness against delays and failures, achieving up to 50.54% faster distribution times compared to state-of-theart methods.