(Untitled)

Dec 17, 2008 09:49

About 16min worth of an interview with some guy who knows a couple things about large filesystems.

His description of a single (temporary?) failed sector triggering a failed disk and subsequent rebuild made me wonder when we'll see a filesystem based on something like erasure codes. Failed read? Rebuild the sector, try a write-read cycle for ( Read more... )

Leave a comment

Comments 2

chezmax December 17 2008, 15:48:46 UTC
That's exactly what nand flash file systems do. ;)

Reply

chezmax December 17 2008, 15:50:47 UTC
Well, nearly. They tend to use a parity code on each sector. NAND tends to have a few bits in a sector fail, so that's recoverable, then they remap the sector.

There's no reason it couldn't be done on an even higher level than that, though.

Reply


Leave a comment

Up