(Information from approximately 2018) part 1: Type read time program time erase time erases per block ---- --------- ------------ ---------- ---------------- SLC 25 uS 200 - 300 uS 1.5 - 2 mS 100,000 Single level cell, one bit per cell MLC 50 uS 600 - 900 uS 3 mS 5,000 Multi ( = two) TLC 75 uS 900 - 1350 uS 4.5 mS 1,000 Three part 2: Random Sequential Device Read Write Read Write ------ ------------- ------------- Samsung 840 Pro SSD 103 287 421 384 Seagate 600 SSD 84 252 424 374 Intel SSD 335 SSD 39 222 344 354 Seagate Savvio 15K.3 HDD 2 2 223 223 (disc for comparison) All in MB/S part 3: LSE = Latent Sector Error URE = Unrecoverable Read Error Error type cheap costly e.g. SATA e.g. SCSI ---------- --------- --------- LSEs 9.4% 1.4% Corruption 0.5% 0.05% (data from 1,500,000 discs over three years) part 4: Findings about LSEs: Costly drives with more than 1 LSE are just as likely to develop additional errors as cheaper drives. For most drives, annual error rate increases in year 2. The number of LSEs increases with disc size. Most discs with LSEs have less than 50. Discs with LSEs are more likely to develop additional LSEs. There is a significant amount of spatial and temporal locality. Disc scrubbing is useful (most LSEs were found this way) Findings about corruption: Chance of corruption varies greatly across different models within the same drive class. Age effects are different across models. Workload and disc size have little impact on corruption. Most discs with corruptions only have a few. There is sptial locality and some temporal locality. There si a weak correlation with LSEs. From "Operating Systems: Three Easy Pieces"