Walrus: An Efficient Decentralized Storage Network

TL;DR

Walrus addresses the fundamental tension in decentralized blob storage between replication overhead, recovery efficiency, and security by introducing Red Stuff, a two-dimensional erasure-coding scheme that enables self-healing with bandwidth proportional to lost data and robust Byzantine fault tolerance in asynchronous networks. Coupled with a novel multi-stage epoch-change protocol and a blockchain-based control plane, Walrus achieves low replication cost ($4.5\times$) and maintains data availability during churn without downtime. The authors provide a full production-grade implementation on public testnet and mainnet, with extensive measurements showing sub-second reads for large blobs, tens of seconds for writes, and scalable epoch transitions. These contributions offer a practical, scalable path to high-integrity decentralized blob storage suitable for a wide range of applications, balancing storage overhead, throughput, and resilience.

Abstract

Decentralized storage systems face a fundamental trade-off between replication overhead, recovery efficiency, and security guarantees. Current approaches either rely on full replication, incurring substantial storage costs, or employ trivial erasure coding schemes that struggle with efficient recovery especially under high storage-node churn. We present Walrus, a novel decentralized blob storage system that addresses these limitations through multiple technical innovations. At the core of Walrus is RedStuff, a two-dimensional erasure coding protocol that achieves high security with only 4.5x replication factor, while enabling self-healing recovery that requires bandwidth proportional to only the lost data $(O(|blob|/n)$ versus $O(|blob|)$ in traditional systems). Crucially, RedStuff is the first protocol to support storage challenges in asynchronous networks, preventing adversaries from exploiting network delays to pass verification without actually storing data. Walrus also introduces a novel multi-stage epoch change protocol that efficiently handles storage node churn while maintaining uninterrupted availability during committee transitions. Our system incorporates authenticated data structures to defend against malicious clients and ensures data consistency throughout storage and retrieval processes. Experimental evaluation demonstrates that Walrus achieves practical performance at scale, making it suitable for a wide range of decentralized applications requiring high-integrity, available blob storage with reasonable overhead.

Figures (7)

• Figure 1: Encoding a Blob in one dimension. First the blob is split into $f+1$ systematic slivers and then a further $2f$ repair slivers are encoded
• Figure 2: 2D Encoding / Red Stuff
• Figure 3: Nodes $1$ and $3$ helping Node $4$ recover its sliver pair
• Figure 4: Walrus write flow. The user generates the blob id of the file they wish to store; acquire storage space through the blockchain; submit the encoded file to Walrus; collect $2f+1$ acknowledgements; and submit them as proof of availability to the blockchain.
• Figure 5: Median read and write latencies for various unencoded blob sizes on Walrus and read latencies on Arweave. Arweave writes require minutes and so are presented in \ref{['tab:arweave-and-filecoin']}. Error bars indicate the 5th and 95th percentile latencies.

Theorems & Definitions (12)

• Definition 1: Complete Data Storage
• Definition 2: Asynchronous Complete Data Storage
• Lemma 1: Primary Sliver Reconstruction
• proof
• Lemma 2: Secondary Sliver Reconstruction
• proof
• Theorem 1
• proof
• Theorem 2
• proof