The NAND flash memories have endurance issues in a high-performance computing (HPC) environment because of their blocks’ limited lifetime. Hence, the NAND flash block allocation and reclamation policies have a significant impact on its lifetime improvement. To address this problem, the scoring-based method is proposed in this paper, which employs priority-based victim block and active block selection methods. The priority of a block is estimated based on both of the score and a time factor. The comprehensive experiments using both realistic and synthetic benchmark traces are performed. The SSDsim simulation environment is extended to implement this method. The results reveal that in comparison with other methods, the proposed method improves endurance by at least 38%.

中文翻译：

---

一种基于评分的基于NAND闪存的固态驱动器受害者块选择的新方法

NAND 闪存在高性能计算 (HPC) 环境中存在耐久性问题，因为它们的块寿命有限。因此，NAND 闪存块分配和回收策略对其寿命改进有重大影响。为了解决这个问题，本文提出了基于评分的方法，它采用了基于优先级的牺牲块和主动块选择方法。块的优先级是基于分数和时间因素来估计的。执行使用现实和合成基准跟踪的综合实验。SSDsim 仿真环境被扩展以实现该方法。结果表明，与其他方法相比，所提出的方法将耐力提高了至少 38%。