Hyperledger Fabric provided by IBM is also a type of alliance-type blockchain, and achieves rapid transaction verification through "PBFT"; a consensus algorithm in which transactions are verified through the consent of most trusted nodes.
Hyperledger Fabric architecture uses a guaranteed publish-subscribe mode messaging channel (such as topic partitioning in Kafka) to separate the consensus service from the transaction log (ledger). The consensus service is provided by a network node called Orderers, and the ledger is managed by the Peer node.
Each Peer node is connected to one or more channels of the consensus service, just like a client in a publish-subscribe communication system. The transactions broadcast on the channel are arranged in the order of consensus (such as PBFT, kafka), and the Peer nodes that subscribe to the channel receive the encrypted blocks. Each peer node verifies the block and submits it to the ledger, and then provides the application with other services that use the ledger.
Multi-channel messaging is supported on the consensus service, so that Peer nodes can subscribe to any number of channels based on application access control policies; that is, the application specifies a subset of Peer nodes to set up channels. These peers constitute the set of stakeholders submitted to the channel for transactions, and only these peers can receive blocks containing related transactions, completely isolated from other transactions.
In addition, a subset of peers submits these private blocks to different ledgers, allowing them to protect these private transactions from the ledgers of other subsets of peers. The application decides to send transactions to one or more channels based on business logic. This is not a built-in limitation, the blockchain network does not know and assumes that there is no relationship between transactions on different channels.
Fabric's multi-channel mechanism is not suitable for public chain systems. Because each node of the public chain may be unreliable, and nodes may be added or deleted at any time, various exceptions and attacks will occur, and a fixed subset of peers cannot be guaranteed.