A Hardware-Assisted Log-Structured File System with Temporal Locality Optimization under Resource Constraints

With the rapid advancement of IoT devices, there is an increasing demand for MPU environments that are low in overall system resources while achieving low power consumption, high energy efficiency, and high performance.Such systems must be capable of controlling multiple sensors and wireless communication modules on minimal battery power, while retaining several days to weeks of logs even under unstable wireless conditions.In logging applications, sensor data are continuously generated and written to non-volatile memory.However, conventional file systems are not designed to handle these operations efficiently. This is because attempting to simultaneously achieve low power consumption, high performance, and power-failure resilience introduces significant overhead and complexity, ultimately degrading energy efficiency.In short, traditional file systems inherently face trade-offs between power efficiency, power-failure resilience functionality, and performance. In typical designs, achieving low power consumption, high performance, and power-failure resilience without relying on specialized modules or custom hardware components requires more resources—such as CPU cycles, power, circuit size, and component count—leading to increased complexity.Therefore, realizing these properties using only general-purpose resources remains a challenging task.Compact/logging type: The proposed time-locality-optimized, hardware-cooperative log-structured file system achieves low power consumption, high performance, and resilience to unexpected power loss using only common components and interfaces, without special or custom hardware.This study reconsiders the boundary design between file system control and hardware-cooperative control in a bare-metal environment without an operating system, and presents a deterministic storage control model based on temporal locality.