Joint decoding of RAID-ECC solutions for SSDs
Ning Zheng, Qing Li · 2017
There are two independent layers of error correction mechanisms, i.e., error-correction codes (ECC) and redundant array of independent disks (RAID), to protect NAND-based Solid State Drives (SSDs) from serious noise/disturb. However, bit error rates of modern SSDs have increased to the point that more powerful ECCs or Maximal Distance Separate codes are only regarded as sufficient solutions. The theme of this work is to show that instead of designing new error correction mechanisms, it is possible to improve the NAND-based SSD reliability by designing a joint decoder exploring the inherent information between ECCs and RAIDs. We first show that it is possible to recover two-page failures using the inherent information with RAID 4/5; we then show that these ideas can be extended to RAID 6; finally, we conduct information theoretic study to explore how much information can be reliably stored with the help of RAID system in NAND-based SSD.