22 Jan
2010
22 Jan
'10
10:14 a.m.
On Thu, 21 Jan 2010 12:44:29 -0800
Larry McVoy <lm(a)bitmover.com> wrote:
Disks are much higher volume and are forced to be
reliable "enough".
That "enough" is the critical point. Check
the average read failure
rate of SATA drives. It is 1 per 10^14 bits for my ST3500320AS. The
drive has a capacity of 512 bytes/sect x 976773168 sectors. I.e. I will
get at least one read error when I try to read the entire disk 25
times. If I build a RAID out of 5 of this drives I only need 5 reads
through the RAID to get a read error and the RAID will degrate.
If you aren't backing up with a crc then you are
doing it wrong.
With a CRC you can detect bit rott. (Probably. Somthing like MD5 or
SHA-1 is mch better then a CRC.) But you can't repair the defect data.
So you want ECC...
--
tschüß,
Jochen
Homepage: http://www.unixag-kl.fh-kl.de/~jkunz/
_______________________________________________
TUHS mailing list
TUHS(a)minnie.tuhs.org
https://minnie.tuhs.org/mailman/listinfo/tuhs