Contents I. Introduction
Blockchain technology has gone from Bitcoin to decentralized, tamper proof, transparent, and traceable distributed database solutions. According to different node access mechanisms, blockchain can be divided into three categories: public blockchain, consortium blockchain, and private blockchain. Among them, the consortium chain has a certain entry threshold, and nodes need authorization to join the blockchain network. It is a decentralized blockchain and an important direction for future blockchain development. The alliance chain currently has numerous applications in data storage [5], copyright management [6], privacy protection [7], data tracing [8], and other areas. In most cases, the alliance chain does not require the design of incentive layers to drive mining and the execution of smart contracts, reducing the cost of system deployment and maintenance; Meanwhile, numerous enterprise blockchain platforms represented by Hyperledger Fabric [9] support pluggable consensus algorithms and multiple languages for writing smart contracts, further promoting the implementation of a large number of applications based on consortium chains. However, the alliance chain still faces the problem of poor scalability. The practical Byzantine Fault Tolerance (PBFT) consensus algorithm, which is currently widely used in consortium chains, does not require a large amount of computing power. However, as the number of nodes increases, its communication overhead will increase in a square order, resulting in lower consensus efficiency. In fields such as finance, smart cities, and supply chain management, high requirements are placed on the performance and scalability of blockchain, typically requiring a single chain to reach a throughput of tens of thousands of levels, the networking ability of thousands of nodes, and transaction confirmation latency of seconds. PBFT algorithms are difficult to meet the needs of these fields [11]. Meanwhile, the network structure of the PBFT algorithm is relatively static, and the entire system must be restarted when dynamically adding nodes [12]. Only improving the PBFT algorithm in references [13] to [24] cannot fundamentally solve the problems existing in the PBFT algorithm, such as single point of failure, data consistency, hardware cost, etc. A sharded blockchain consensus algorithm K-RPBF with supervised nodes was proposed in reference [25], which uses an improved Raft algorithm for consensus within shards and PBFT algorithm for consensus among Raft leaders in each shard. Adopting sharding and multi center approaches goes against the decentralized approach of blockchain, and consensus is required between multiple centers, while consensus is required within one center, which requires two types of consensus processes and increases system overhead. The Raft algorithm is a fault-tolerant algorithm designed for faulty nodes rather than malicious ones. The K-RPBF consensus algorithm mainly consists of node detection, initial sharding, resharding, intra shard consensus, and supervision, as shown in Figure 1.
Fragmented blockchain model
