docs/content/concepts/architecture.md

title	description
Architecture Overview	Simplicity and Interoperability via a DataNFT Core

Overview

Here is the Ocean architecture.

Here’s an overview of the figure.

• The top layer is applications like Ocean Market. With these apps, users can onboard services like data, alogrithm, compute-to-data into crypto (publish and mint DataNFTs and Datatokens), hold datatokens as assets (data wallets), discover assets and buy / sell datatokens for fixed or auto-determined price (data marketplaces), and consume data services (consume datatokens).
• Below that are libraries used by the applications: Ocean.js (JavaScript library) and Ocean.py (Python library). This also includes middleware to assist discovery:
  • Aquarius: Provides metadata cache for faster serach by caching on-chain data into elasticsearch
  • Provider: Facilitates downloading assets, DDO encryption and communicating with operater-service for Compute-to-Data jobs.
  • TheGraph: 3rd party tool Developers can utilize the libraries to built thier custom applications and marketplaces.
• The lowest level has the smart contracts used by the libraries. They’re deployed on Ethereum mainnet, and other compatible networks. To see the list of supported networks click here.

Left to right are groupings of functionality: tools for datatokens, tools for markets (including pools), tools to consume data services and for metadata, and external ERC20, ERC721 tools.

The rest of this page elaborates.

DataNFT

DataNFTs are based on ERC721 standard. The publisher can use Marketplace or client libraries to deploy a new DataNFT contract. To save gas fees, it uses ERC1167 proxy approach on the ERC721 template. Publisher can then assign manager role to other ethereum addresses who can deploy new Datatoken contracts and even mint them. Each Datatoken contract is associated with one DataNFT contract. Click here to further read about DataNFTs and Datatokens.

Datatokens & Access Control Tools

The asset can be a dataset or an algorithm. The publisher actor holds the asset in Google Drive, Dropbox, AWS S3, on their phone, on their home server, etc. The publisher can optionally use IPFS for a content-addressable URL. Or instead of a file, the publisher may run a compute-to-data service.

In the publish step, the publisher invokes Ocean Datatoken Factory to deploy a new datatoken to the chain. To save gas fees, it uses ERC1167 proxy approach on the ERC20 datatoken template. The publisher then mints datatokens.

The publisher runs their own Ocean Provider or can use one deployed by Ocean Protocol. In the consume step, Provider software needs to retrieve the data service URL given a datatoken address. One approach would be for the publisher to run a database; however this adds another dependency. To avoid this, Provider encrypts the URL and it URL on-chain.

To initiate the consume step, the data consumer sends 1.0 datatokens to the Provider wallet. Then they make a service request to the Provider. The Provider loads the encrypted URL, decrypts it, and provisions the requested service (send static data, or enable a compute-to-data job).

Instead of running a Provider themselves, the publisher can have a 3rd party like Ocean Market run it. While more convenient, it means that the 3rd party has custody of the private encryption/decryption key (more centralized). Ocean will support more service types and url custody options in the future.

Ocean JavaScript and Python libraries act as drivers for the lower-level contracts. Each library integrates with Ocean Provider to provision & consume data services, and Ocean Aquarius for metadata. Ocean React hooks use the JavaScript library, to help build web apps & React Native apps with Ocean.

Market Tools

Once someone has generated datatokens, they can be used in any ERC20 exchange, centralized or decentralized. In addition, Ocean provides a convenient default marketplace that is tuned for data: Ocean Market. It’s a vendor-neutral reference data marketplace for use by the Ocean community.

The marketplaces are decentralized (no single owner or controller), and non-custodial (only the data owner holds the keys for the datatokens).

Ocean Market supports fixed pricing and automatic price discovery.

• For fixed pricing, there’s a simple contract for users to buy/sell datatokens for OCEAN, while avoiding custodianship during value transfer.
• For automatic price discovery, Ocean Market uses automated market makers (AMMs) powered by Balancer. Each pool is a datatoken-OCEAN pair. In the Ocean Market GUI, the user adds liquidity then invokes pool creation; the GUI’s React code calls the Ocean JavaScript library, which calls the Pool Factory to deploy a Pool contract. (The Python library also does this.) Deploying a datatoken pool can be viewed as an “Initial Data Offering” (IDO).

Complementary to Ocean Market, Ocean has reference code to ease building third-party data marketplaces, such as for logistics (dexFreight data marketplace) or mobility (Daimler).

This post elaborates on Ocean marketplace tools.

Metadata Tools

Metadata (name of dataset, date created etc.) is used by marketplaces for data asset discovery. Each data asset can have a decentralized identifier (DID) that resolves to a DID document (DDO) for associated metadata. The DDO is essentially JSON filling in metadata fields. For more details on working with OCEAN DIDs check out the DID concept documentation. The DDO Metadata documentation goes into more depth regarding metadata structure.

OEP8 specifies Ocean metadata schema, including fields that must be filled. It’s based on the public DataSet schema from schema.org.

Ocean uses the Ethereum mainnet and other compatible networks as an on-chain metadata store, i.e. to store both DID and DDO. This means that once the write fee is paid, there are no further expenses or dev-ops work needed to ensure metadata availability into the future, aiding in the discoverability of data assets. It also simplifies integration with the rest of the Ocean system, which is Ethereum-based. Storage cost on Ethereum mainnet is not negligible, but not prohibitive and the other benefits are currently worth the tradeoff compared to alternatives.

Due to the permissionless, decentralized nature of data on Ethereum mainnet, any last-mile tool can access metadata. Ocean Aquarius supports different metadata fields for each different Ocean-based marketplace. Developers could also use TheGraph to see metadata fields that are common across all marketplaces.

Third-Party ERC20 Apps & Tools

The ERC20 nature of datatokens eases composability with other Ethereum tools and apps, including MetaMask and Trezor as data wallets, DEXes as data exchanges, and more. This post has details.

Actor Identities

Actors like data providers and consumers have Ethereum addresses, aka web3 accounts. These are managed by crypto wallets, as one would expect. For most use cases, this is all that’s needed. There are cases where the Ocean community could layer on protocols like Verifiable Credentials or tools like 3Box.