Existing decentralized storage protocols fall short of the service required by real-world applications. Their throughput, latency, cost-effectiveness, and availability are insufficient for demanding workloads such as video streaming, large-scale data analytics, or AI training. As a result, Web3 data-intensive applications are predominantly dependent on centralized infrastructure. Shelby is a high-performance decentralized storage protocol designed to meet demanding needs. It achieves fast, reliable access to large volumes of data while preserving decentralization guarantees. The architecture reflects lessons from Web2 systems: it separates control and data planes, uses erasure coding with low replication overhead and minimal repair bandwidth, and operates over a dedicated backbone connecting RPC and storage nodes. Reads are paid, which incentivizes good performance. Shelby also introduces a novel auditing protocol that provides strong cryptoeconomic guarantees without compromising performance, a common limitation of other decentralized solutions. The result is a decentralized system that brings Web2-grade performance to production-scale, read-intensive Web3 applications.
@article{arxiv.2506.19233,
title = {Shelby: Decentralized Storage Designed to Serve},
author = {Guy Goren and Andrew Hariri and Timothy D. R. Hartley and Ravi Kappiyoor and Alexander Spiegelman and David Zmick},
journal= {arXiv preprint arXiv:2506.19233},
year = {2025}
}