Method for Adaptive Threshold Start Relocation Driven Garbage Collection Scheme for NAND Flash Memory-based Solid State Drives (SSDs)

Abstract

In NAND flash memory, host-initiated rewrites and deletions can fragment data within blocks, while wear-out factors further challenge data integrity and drive lifespan. Relocation, a data consolidation or migration process, is crucial for maintaining NAND performance and endurance. A Solid-State Drive (SSD) storage device enhances writing performance through an automated process called garbage collection, where a whole block erasure is optimally avoided before each write operation. Most SSD controllers execute garbage collection algorithms during off-peak hours to preserve optimal write rates during regular operations. To prevent overused blocks from wearing out, most SSD controllers include wear leveling in their garbage collection processes. This allows P/E cycles to be distributed more evenly throughout the storage blocks. However, conventional fixed-threshold relocation algorithms for Garbage Collection (GC), particularly in consumer NAND products with constrained DRAM, often need to optimize NAND capabilities fully. To address this, we propose an innovative adaptive relocation threshold algorithm that dynamically adjusts relocation initiation based on real-time NAND health metrics, including Program/Erase Cycles (PEC) and Write Amplification (WA). This algorithm delivers significantly smoother SSD performance, extends product lifespan, and meticulously balances NAND characteristics, unlocking its full potential.