How storage works (high-level steps): When you store a file, your client:
Encode & Commit: Chops the file into slivers via Red Stuff and computes cryptographic commitments for each sliver and for the overall blob.
Register on Sui: Submits a Sui transaction to reserve space, recording the blob’s commitment and paying the fee in WAL.
Distribute Fragments: Sends each chosen storage node its assigned primary and secondary sliver, along with the commitments for verification.
Node Verification: Each node checks that the sliver it received matches the on-chain commitment. Honest nodes sign an acknowledgment if correct.
Publish Proof: The client collects enough acknowledgments (a quorum) and publishes a Proof of Availability (PoA) certificate on Sui. This on-chain certificate is a permanent record that a sufficient set of nodes has the data and will keep it for the paid duration.
By representing each blob as a Sui object and using Red Stuff’s efficient coding, Walrus achieves high resilience with only ~4–5× replication overhead, instead of tens or hundreds of copies. This means the data remains recoverable even under heavy node churn, while keeping storage and recovery costs low.