5

I understand from NEC's white paper, Silent data corruption in disk arrays, that:

  • some SAS drives should have a "T10-DIF" feature to detect silent data corruption; whereas
  • "There is no standard for ATA-based drives (including SATA) that protects against silent data corruption at the SCSI level in the storage technology stack."

The point of that white paper is to inform people of NEC's proprietary technology for protecting against silent data corruption in SATA drives. However, ZFS seems to provide at least equivalent protection, and is preferable to me as it is not proprietary (except for Oracle's most recent ZFS revisions).

I have two questions:

  1. Am I right in thinking that using ZFS with T10-DIF SAS drives would give an additional layer of protection against silent data corruption as compared to using just one of those two technologies alone?

  2. Given that T10-DIF SAS drives do not seem to be readily available, what reasons are there - if any - to prefer non-T10-DIF SAS drives over equivalent* SATA drives, or vice versa?

* By this, I mean comparing like with like: e.g. enterprise class SAS drives from a given manufacturer vs enterprise class SATA drives from the same manufacturer.

Improve this question
1
  • 1
    Having the array manager check the disks for corruption on a regular basis is a tried and true method of preventing/detecting silent data corruption. ZFS and many other managers call it "Scrubbing". It should be done regularly, and results reported if abnormal (most frequently for critical data, but at least monthly as a dead minimum). Great little writeup on what's going on: how multi-disk failures happen.
    – Chris S
    Dec 2, 2013 at 14:57

1 Answer 1

Reset to default
3

If you can avoid SATA disks, please do so. Nearline SAS is a good compromise for people using ZFS who want stability, performance and capacity. You don't need proprietary silent data corruption technologies from the disk manufacturers...

The arguments against SATA really come down to the implementation and choice of storage hardware, controllers, enclosure, etc. It's much easier to go wrong or to find a particularly toxic combination when SATA is in the mix; e.g. behind an expander, using bad controllers and in various failure modes.

See: How can a single disk in a hardware SATA RAID-10 array bring the entire array to a screeching halt?

Improve this answer

You must log in to answer this question.