RAIN

For engineers · sourced from the 2017 paper

How RAIN works.

The machinery behind the plain-English story: erasure coding, villages, ledger-backed monitoring, and distributed self-repair — as specified by Monti & Rasmussen in Artificial Life 23(4). Where this page says a number, the paper said it first. Prefer plain words? Start at What is RAIN.

01 · Erasure coding

A file becomes K = r·N blocks. Any N rebuild it.

Files are encrypted on the owner's device, then erasure-coded into K blocks such that any N of them reconstruct the original — Reed–Solomon-style polynomial codes, the same family that protects optical media. The paper's working redundancy ratio is r = 1.5: a file kept as twelve blocks survives the loss of any four. No block alone reveals anything; the encryption happens before coding, so a node stores noise.

Redundancy ratio (paper)
r = 1.5
Worked example
12 blocks, any 8 rebuild
Code family
Reed–Solomon erasure codes
Encryption
Client-side, before anything leaves the device
Fig. — Twelve erasure-coded blocks at r = 1.5. Scroll: four fail; eight remain; the file rebuilds. Schematic.

02 · Villages

Blocks live in a village of K nodes that trade storage.

A village is a group of K nodes — inexpensive, always-on devices like a Raspberry Pi with a flash drive — that store one another's blocks in exchange for having their own stored. The paper models each node contributing a fixed share, Z = 100 GB. Every node operates on local interactions with a limited set of peers it does not need to trust a priori; there is no coordinator, no central index, and no single point whose failure matters.

Village size
K nodes
Storage share (paper model)
Z = 100 GB per node
Trust required between peers
None a priori
Central point of failure
None

03 · Monitoring & self-repair

Merkle-tree monitoring on a village ledger. Repair is automatic.

Nodes continuously audit one another's availability and integrity through Merkle-tree hashes recorded on a ledger the village shares. Because the village is small and its nodes are always on, this ledger is local — kept consistent among peers without global consensus, secured with proofs of space over storage the nodes have already committed, not with mined work. When the number of live blocks for a file falls toward the threshold, the surviving nodes regenerate the missing blocks from what remains and place them on fresh nodes. A tampered block fails its hash, is discarded, and is rebuilt the same way.

Integrity checks
Merkle-tree hashes, continuous
Record of checks
Village-local ledger
Sybil / commitment defense
Proofs of space (no mining)
Repair
Distributed regeneration onto fresh nodes

04 · What the model yields

36 nodes

give a stored file an expected lifetime on the order of the age of the Earth.

The paper's reliability model, with home-grade hardware and realistic failure rates: redundancy plus continuous repair compounds, so village lifetime grows explosively with size. Thirty-six ordinary nodes is enough to put a file's expected survival in the billions of years. Not because any machine lasts — because the village never stops replacing what fails.

05 · Against the alternatives

Engineered for hardware you cannot trust.

Against the cloud
A data center concentrates custody: one operator, one key-holder, one legal and physical target — and more than 3% of the world's power consumption feeds the model. RAIN's operator does not exist: keys never leave the owner, and custody is the village.
Against storage-trading networks
Systems in the Storj lineage assume well-connected, reliable machines. RAIN is engineered for the opposite — home links and drives that fail — and pulls its reliability from erasure coding plus self-repair, not from the hardware underneath.
Against blockchains
A global chain needs asynchronous worldwide consensus, so it pays in proof-of-work and latency. A village is small, always-on, and local: its ledger stays consistent among peers, secured by proofs of space over storage already committed. No mining, no token, no dedicated hardware.

Sources

Check every claim.

Primary · 2017

RAIN: A Bio-Inspired Communication and Data Storage Infrastructure

Monti & Rasmussen. Artificial Life 23(4), 552–557. MIT Press. Every parameter on this page — r, K, Z, the 36-node lifetime — is theirs.

Read the paper

The full reference list — Reed & Solomon 1960, Merkle 1988, proofs of space, the working papers — lives on the reading list.

Plain-English version: What is RAIN.