A Three-Tier Local-Group-Offchain Architecture for Blockchain Storage Optimization

To ensure data security, blockchain systems inherently rely on a full-node replication mechanism that mandates each node to store the entire ledger. While this approach guarantees tamper resistance and traceability, it imposes substantial storage overhead on nodes and severely limits scalability. Existing storage optimization strategies primarily fall into two categories: on-chain methods (e.g., erasure-coded grouped storage using Reed-Solomon codes) and off-chain solutions (e.g., migrating partial data to external systems like the InterPlanetary File System [IPFS]). However, these approaches introduce a critical trade-off between node storage overhead reduction and data access performance degradation, necessitating further equilibrium optimization in both dimensions. To address this challenge, we propose a three-tier "Local-Group-Offchain" storage architecture that dynamically categorizes blocks into distinct storage modes based on their predicted access probabilities: (1) high-access-probability blocks (hot blocks) retain full local copies for low-latency access, (2) medium-access blocks (warm blocks) utilize Reed-Solomon (RS) erasure coding for space-efficient grouped storage, and (3) low-access blocks (cold blocks) migrate to IPFS with Merkle-proof verification. A random forest algorithm is employed to predict block access probabilities, enabling adaptive classification. Theoretical analysis and experimental evaluations demonstrate that the proposed storage optimization scheme achieves a good balance between node storage cost and data access performance compared to previous schemes, offering superior scalability and usability.