NAND retention relaxation
NAND retention relaxation trades a shorter cell-level data-holding requirement for another operating benefit while separately protecting data that must survive longer.
One design trades programming precision or ECC strength for faster writes, then tracks each block's remaining retention time. Before that interval runs out, surviving valid data are moved to another block. Frequent overwrites are an opportunity, not a substitute for tracking the data that remain. [1]
A different objective is greater erase-cycle endurance at a fixed ECC capability, supported by timely correction and refresh. These are different trade-offs: a write-latency result and an endurance result are not interchangeable. [2]
Refresh requires powered operation. A policy that renews data while a device runs does not, by itself, extend the safe duration of an unpowered archive. [2]
Relations
Limitations
- The cited mechanisms depend on their NAND model, coding assumptions and maintenance policy; they are not a host setting that safely shortens every SSD's retention requirement. [1, 2]
Citations
- Liu, Yang and Wu. Optimizing NAND Flash-Based SSDs via Retention Relaxation. FAST 2012, sections 3, 5.1 and 6.
- Cai et al. Flash Correct-and-Refresh. ICCD 2012, sections III.C and IV. DOI: 10.1109/ICCD.2012.6378623.
Sources & caveats
- Primary-paper synthesis of distinct historical designs; no universal retention period, device command or measured speedup is claimed.