Enhancing reliability of 3D portable SSDs by storing redundant data to RRAM cache considering layer-to-layer process variation
摘要
Solid-state drives (SSDs) are extensively used in various storage scenarios due to their superior device characteristics. However, 3D portable SSDs requiring long-term data storage after power-off face significant data reliability issues stemming from retention loss and layer-to-layer process variation. Data stored on 3D portable SSDs may not be read correctly within just a few days after power-off, leading to severe consequences such as potential loss of critical personal and business information, as well as increased recovery costs. In this paper, we propose a novel RAID scheme called weak-layer-priority and interleaved RAID (WLPI-RAID), which leverages resistive random access memory (RRAM) cache to store redundant parity data after power-off. Therefore, WLPI-RAID effectively addresses the long-term data retention loss problem without incurring additional storage overhead and with minimal impact on normal read/write operations. WLPI-RAID employs a weak-layer-priority strategy to significantly reduce RAID overhead for parity storage, computation, and read/write operations. Furthermore, WLPI-RAID constructs RAID groups in an interleaved manner to prevent early failure of weak layers. Our approach substantially enhances SSD reliability, extending data retention time by 8.1 times compared to solutions that only adopt ECC error correction code (ECC), lasting over a year.