LazyDB: I/O Optimization of Blockchain Storage System Based on Non-Volatile Memory

For the blockchain storage system, this paper proposes an I/O optimization scheme based on non-volatile memory based on the storage performance requirements of the blockchain system application. This scheme adopts the key-value separation strategy. Due to the uncertain size of transaction data, by separating the “key” and “value” of transaction data, the inherent read-write amplification problem of the LSM-Tree structure is solved. The key-value separation strategy redesigns the storage system's space management and garbage collection mechanisms. Therefore, this chapter uses the byte-addressing characteristics of non-volatile memory to build a new data storage structure, a lazy compression mechanism, and a lazy garbage collection mechanism. On this basis, this paper builds a LazyDB database system for blockchain storage. To ensure the safe and reliable operation of the blockchain system, it can make full use of the storage medium and data characteristics, reducing the blockchain reading and writing process. The I/O overhead in the storage system improves the storage performance of the system. The experimental results show that compared with the most commonly used databases, such as LevelDB and RocksDB in the blockchain system, LazyDB effectively reduces the write amplification of the system and has higher read, write, and data update performance.