Design and Research on Full-Memory Storage Scheme for Relational Databases
Hong Lin Wu · 2025
With rising demands for high-speed data processing, traditional disk-based RDBMSs hit I/O limits. We propose a full-memory relational engine that keeps all data in RAM yet preserves ACID via efficient logging and periodic checkpoints. Its design comprises storage conversion modes, a multi-layer middleware architecture, and synchronized persistence strategies, all detailed with system and data-model diagrams. On standard hardware, it achieves up to $5 \times$ higher transaction throughput and lower latency compared to disk-only deployments, while maintaining durability. This blueprint demonstrates how to balance performance, cost, and data safety in high-speed in-memory relational systems.