Noms — The Decentralized Database
Noms makes it
easy tractable to create rich,
multiuser, collaborative, fully-decentralized applications.
Like most databases, Noms features a rich data model, atomic transactions, support for large-scale data, and efficient searches, scans, reads, and updates.
Unlike any other database, Noms has built-in multiparty sync and conflict resolution. This feature makes Noms a very good fit for P2P decentralized applications.
Any number of dapp peers in a P2P network can concurrently modify the same logical Noms database, and continuously and efficiently sync their changes with each other. All peers will converge to the same state.
For many applications, peers can store an entire local copy of the data they are interested in. For larger applications, it should be possible to back Noms by a decentralized blockstore like IPFS, Swarm, or Sia (or in the future, Filecoin), and store large-scale data in a completely decentralized way, without replicating it on every node. Noms also has a blockstore for S3, which is ideal for applications that have some centralized components.
We'd love to talk to you about the possibility of using noms in your project so please don't hestitate to contact us at email@example.com.
How it Works
Think of Noms like a programmable Git: changes are bundled as commits which reference previous states of the database. Apps pull changes from peers and merge them using a principled set of APIs and strategies. Except that rather than users manually pulling and merging, applications typically do this continuously, automatically converging to a shared state.
Some additional features include:
- Versioning: It’s easy to use, compare, or revert to older database versions
- Efficient diffs: diffing even huge datasets is efficient due to noms’ use of a novel BTree-like data structure called a Prolly Tree
- Efficient storage: data are chunked and content-addressable, so there is exactly one copy of each chunk in the database, shared by other data that reference it. Small changes to massive data structures always result in small operations.
- Verifiable: The entire database rolls up to a single 20-byte hash that uniquely represents the database at that moment - anyone can verify that a particular database hashes to the same value
Read the Noms design overview.
For overall status of the database, see Noms Status.
For the decentralized use case in particular: we are fairly confident in this approach and are actively looking for partners to work with to build it out.
- Demonstrate core concept of using Noms to continuously sync across many users (Done! See noms-chat demos)
- Demonstrate using libp2p or similar to traverse NATs
- Investigate backing IPFS with Noms rather than the reverse - this should improve stability and dramatically improve local performance
- Demonstrate using IPFS with a schema that permits nodes to disappear