Chapter 3 — Architectural Fundamentals:
3.1 Blob Storage vs. Standard Blockchain Data
Blockchains are excellent at storing small, transaction level data. They struggle with large files because:
Storing every byte on chain is costly.
Network performance is degraded with huge on chain storage.
Walrus uses a blob storage model:
📌 Files are uploaded as “blobs” essentially binaries that represent unstructured data (videos, images, AI models).
📌 Metadata and proofs of storage are stored on Sui, while the actual data is distributed across storage nodes.
This model reduces onchain bloat while still preserving verifiability.
3.2 Red Stuff Advanced Erasure Coding
Central to Walrus’s performance is its Red Stuff erasure coding algorithm:
🔹 Red Stuff splits a blob into fragments (“slivers”),
🔹 Adds redundancy using advanced coding techniques,
🔹 Stores these fragments across many nodes.
The key benefit: even if a significant number of nodes go offline up to about 66% the original file can still be reconstructed using remaining slivers. This provides strong fault tolerance with far less replication overhead than naive copying.
3.3 Delegated Proof-of-Stake (dPoS)
Walrus uses a delegated proof-of-stake consensus model:
Token holders can delegate WAL to trusted node operators.
Validators & storage nodes are selected to serve content and participate in network operations.
Rewards and penalties balance incentives to ensure nodes remain honest and available.
This approach provides scalability and security while enabling token holders to participate even without operating a node directly.
