Design and Implementation of Distributed Price Oracle Machine System

To avoid single points of failure, some Oracle Machine services adopt a distributed design. Taking the service that provides the BTC to USD price as an example, it aggregates data from 31 independent price Oracle Machines to provide information to users.

The contract of this aggregator can be viewed on the blockchain explorer. By calling specific methods in the contract, all off-chain Oracle Machine information contained in the aggregator can be obtained. Each off-chain Oracle Machine can provide price data in response to user requests in the aggregator by calling designated methods. These off-chain Oracle Machines are usually external accounts that not only provide data for the BTC/USD aggregator but may also provide price information for other aggregators such as ETH/USD.

The processing flow of on-chain contracts includes the following steps:

1. Read the current contract status and perform a series of checks.
2. Make necessary preparations
3. Verify each signature data, checking the role of the signer and the uniqueness of the signature.
4. Sort the observations, select the median and perform threshold checks.
5. Record the results of this Oracle Machine and conduct final verification.

The final step of verification usually calls methods from other contracts to compare whether the price deviations given by different Oracle Machines are within an acceptable range.

Feed Registry System

To further simplify usage and improve flexibility, some Oracle Machine services have introduced the concept of Feed Registry. This can be understood as an aggregator of multiple price information sources, allowing users to directly read price data through the Feed Registry without having to set up each token's price information source themselves.

Multi-level Data Aggregation Mechanism

High-quality price Oracle Machines typically employ a multi-level data aggregation mechanism, mainly consisting of three levels:

1. Data Source Aggregation: Collect raw price data from major trading platforms and perform weighted calculations.

2. Node operator aggregation: Each node obtains prices from multiple independent data sources and takes the median to exclude outliers.

3. Oracle Machine network aggregation: After a preset number of nodes respond, the median of all data is taken again.

The final price updates usually need to meet specific trigger conditions, such as deviation thresholds or heartbeat thresholds. While this mechanism increases security, it also results in relatively slow price updates, making it mainly suitable for application scenarios that are not very sensitive to price updates.