Post AiZTe5F04oFGbT6NsG by dave_andersen@hachyderm.io
 (DIR) More posts by dave_andersen@hachyderm.io
 (DIR) Post #AiZTe1RcBxxWpQyOyu by dave_andersen@hachyderm.io
       2024-06-03T22:50:28Z
       
       1 likes, 0 repeats
       
       As part of repairing a disk failure, I'm replacing a single working 18TB disk with a new 22TB disk, and then re-using the 18TB disk to replace a dead 18TB disk in the big 11-disk RAIDZ-2 array.Every time I do this after another year or three has passed, I appreciate Bianca and Garth's message about the danger of HD capacity increasing quadratically vs transfer speed increasing linearly...1+ day just to copy the disk. (!)(Ref: https://www.cs.cmu.edu/~bianca/fast07.pdf"Disk Failures In The Real World")
       
 (DIR) Post #AiZTe5F04oFGbT6NsG by dave_andersen@hachyderm.io
       2024-06-03T23:01:30Z
       
       0 likes, 0 repeats
       
       By way of explanation: Disks are effectively 2D -- it's a spinning plate. So when you do the magic shrinky thing, you shrink in both the X and Y dimensions. If you shrink one of those dimensions by half, you overall fit 4x more data on.But disks are read linearly - the read head reads one track at a time. So that same 1/2 shrink doubles the read speed but 4x's the storage, so your time to copy the entire disk effectively doubles every time you get 4x the capacity.#storage #disk #raid
       
 (DIR) Post #AiZTe8QoDMTQUxeMCG by dave_andersen@hachyderm.io
       2024-06-03T23:02:43Z
       
       0 likes, 0 repeats
       
       This is a little problematic because it means your window during which you might experience _another_ failure grows. Decades ago, you might have had a 40 minute "double failure will kill you" window; today it's a day. So we've generally had to move to more redundant systems, like the aforementioned RAIDZ-2, which can tolerate two simultaneous failures without data loss.