EDGEHYDRA: AN EFFICIENT FAULT-TOLERANT EDGE DATA DISTRIBUTION FRAMEWORK USING ERASURE CODING

Abstract

Low-latency access in edge computing systems depends on the effective distribution of popular data from the cloud to edge servers. Current distribution systems split a file into many data blocks and send them to target edge servers in simultaneously. Once all of the blocks are received, the file is reconstructed. This method speeds up data delivery, but as the successful reconstruction depends on receiving each data block, it is susceptible to transmission delays and errors brought on by network fluctuations or server outages. This study presents EdgeHydra, an erasure coding-based fault-tolerant data distribution strategy, to overcome these constraints. By encoding a file into both data and parity blocks, EdgeHydra enables edge servers to reconstruct the original file from any adequate subset of these blocks without having to wait for all communications to finish. Additionally, it uses a leaderless block supplement mechanism that allows edge servers to coordinate distributedly to recover lost blocks without the need for a centralized controller. According to experimental data, EdgeHydra achieves up to 50.44% quicker distribution times than state-of-the-art techniques, greatly enhancing resilience against delays and failures.