[Discuss] SSD drives vs. Mechanical drives

[blu at nedharvey.com (Edward Ned Harvey (blu))]

> From: Dan Ritter [mailto:dsr at randomstring.org]
> 
> 3Ware tw_cli man page:

I believe that hardware data integrity checking must exist, so you are correct to call me out on the generalization, "hardware raid doesn't do integrity."  I was in fact overstating my belief.  But I've never seen any hardware that did data integrity, and the example you posted of the 3Ware card, doesn't seem to do it either - The text you posted says the raid card will compare both sides to make sure they're exact, but it doesn't say what it does when it detects a difference between the redundant copies.  Does it mark blocks as corrupt?  Does it have a checksum stored somewhere to determine which conflicted side has the correct data?  Does it constantly checksum all data it reads under normal operation, or does it just compare the two sides when you tell it to compare the two sides?  In order for the hardware to do data integrity, it requires validating all data on every read.

In any event, the info posted is insufficient to believe the hardware provides integrity.  But I'm quite sure if you search for it, *some* hardware out there probably does the job.


> I find your exasperated tone exasperating, by the way.

Fair.  And sorry about that.  But I felt it was justified at the time, because:  

Somebody said raid & mirroring doesn't give you integrity.  I said it does, if you're using btrfs or zfs.  Somebody replied to directly contradict that btrfs and zfs provide integrity, and asserted information that demonstrates ignorance of how zfs and btrfs work.  So I said "Seriously dude?"  And explained how it actually works.  I would have used a different tone, if there had been interrogation, "Really, how can zfs/btrfs do that?"  But instead, it was a negative assertion, contradicting what I just said.  It was a "Duty Calls" moment.   http://xkcd.com/386/ 

The later points in the thread, about how data corruption can reach disk, if the corruption originated in RAM or CPU, or the SAS/SATA/whatever bus, are all fair points.  ZFS and BTRFS are not magic bullets that prevent all forms of data corruption at every layer.  But this is a thread about SSD vs HDD, where the reliability of an individual storage device has been called into question (both SSD and HDD), and mirroring/redundant copies posed as the counter to corruption caused by the storage device.  So with regard to the *storage device* reliability, zfs/btrfs mirroring/redundant copies of data do indeed provide data integrity.