Abstract: Most consortium blockchains now run in closed and deterministic environments, and their smart contracts cannot have IO operations with the outside world. Some application scenarios (such as crediting blockchain, carbon trading blockchain, supply chain, express tracking, etc.) require a mechanism responsible for data interaction with the outside of consortium blockchains, generally called an oracle machine. The existing oracle techniques in the consortium chain have the following shortcomings: 1) The limited data interaction mode cannot meet the needs of distributed applications; 2) With the increase in the number of distributed oracle nodes, the consensus delay will also increase. 3) The participants of the consortium blockchain usually maintain the oracle nodes in the distributed oracle system, and the behavior in the data consensus process is invisible to the blockchain, which is not conducive to data governance. To address the problems, this paper proposes the following methods: 1) Based on the eventdriven mechanism, four oracle design patterns or interaction patterns are proposed, which support Pull and Push, Inbound and Outbound, four combinations of the oracle data interactions; 2) The threshold signature algorithm is used to reach a consensus on the data, which improves the scalability of the oracle system while ensuring the credibility of the data; 3) A reputation mechanism is introduced for data governance to maintain a local and global reputation for each oracle node, and dynamic update is carried out in the data consensus process. Finally, by designing multichain scenairos in crediting blockchain and carbon trading blockchain, the applications of the four oracle design patterns, scalability, and reliability of the oracle nodes are evaluated and analyzed.

Key words: distributed oracle machine, smart contract, oracle machine design pattern, event type, data governance

摘要： 目前许多联盟链都处于封闭、确定性的环境下，其智能合约不能与外界发生IO.一些应用场景(如征信链、碳交易链、供应链、快递追踪等)需要一种机制与联盟链外部进行数据交互，一般被称为预言机 (oracle machine).现有联盟链中的预言机技术存在如下不足： 1)有限的数据交互模式，无法满足分布式应用需求； 2)随着分布式预言机节点数量的增加，共识时延也会随之提升，无法对系统进行有效扩展； 3)分布式预言机系统中的预言机节点通常由联盟链的参与方维护，数据共识过程中的行为对联盟链是不可见的，不利于数据的治理.针对上述问题，提出如下方法： 1)基于事件驱动机制，提出4种预言机设计模式或交互模式，支持入链(Inbound)和出链(Outbound)，以及Pull和Push这4种预言机数据交互组合方式； 2)使用门限签名算法对数据达成共识，在保证数据可信的同时提高预言机系统的可扩展性； 3)引入一种信誉机制进行数据治理，每个预言机节点维护局部信誉和全局信誉，对预言机节点形成监管，提高整个预言机集群的可信度.最后，通过设计实现征信链及碳交易链场景下的多链应用，对4种预言机的设计模式、可扩展性及预言机节点可信度进行评测与分析.

关键词: 分布式预言机, 智能合约, 预言机设计模式, 事件类型, 数据治理