Blockchain Storage Optimization Based on Erasure Code

Blockchain has the characteristics of decentralization, unforgeability, traceability, transparency, etc. Therefore, blockchain can solve the problem of mutual distrust between nodes in the decentralized network and provide a viable way to build a value-connected platform. However, blockchain requires each node to store a complete copy of data to ensure data reliability with high storage redundancy. Unfortunately, it also brings huge storage pressure to blockchain nodes and reduces the utilization of storage resources, which makes the storage scalability be a bottleneck of blockchain. Meanwhile, with the increasing number of transactions, the size of blockchain also increases rapidly, which hinders the nodes with limited storage capacity to join the blockchain system and hence weakens the decentralization of the system. Using erasure code to encode the data stored in the blockchain can effectively reduce the storage redundancy. Nevertheless, the reduction of storage redundancy will reduce the reliability of the data, incur data recovery costs, and increase the data access delay. The existing work on reducing the storage redundancy in blockchains ignores the impact of inter-node delay and encoded data chunk storage locations on data reliability and data access delay during the data chunk placements, while there is a tradeoff between data storage cost and data access latency. In this paper, we study the decision problem of the number and locations of encoded data chunk copies with the erasure code in BFT permissioned blockchains, with the aim to achieve a good balance between data storage cost and data access latency under the data reliability constraint. We also propose a Latency-aware Encoded data chunks Allocation algorithm (LEA). Firstly, algorithm LEA solves the relaxation problem of the studied problem and the dual problem of the relaxation problem. The relaxation simplifies the problem model and facilitates the solution of the problem. We classify the solutions of the relaxation problem into clusters. Specifically, we select a node as the cluster center based on the optimal solution of the dual problem, and put the nodes into different clusters according to the optimal solution of the relaxation problem. Secondly, at least one node in each cluster is randomly selected to store the copy of the coded data chunk according to the optimal solution of the relaxation problem. Finally, the algorithm adjusts the storage allocation solution to satisfy the constraints of the decision problem. Theoretical analysis proves that algorithm LEA is a ln3+2 approximation algorithm to the data allocation decision problem. We conduct extensive experiments in both the simulation environment and the real permissioned blockchain system. The experimental results show that algorithm LEA can achieve good performance in terms of the tradeoff between data storage cost and data access delay, while satisfying the data reliability constraint. Algorithm LEA can effectively reduce the blockchain storage cost, enhance the blockchain storage capacity with the increase of the number of nodes, and realize the horizontal scalability of the system storage. Compared with the existing algorithms, algorithm LEA can achieve better data access performance in different networks, and the performance improvement can reach 15.2%. In addition, we evaluate the impact of the number of malicious nodes on the performance of the algorithm. When there are malicious nodes in the network, algorithm LEA can still obtain favorable data access performance while ensuring data reliability, which outperforms the existing algorithms by 15.4%. © 2022, Science Press. All right reserved.