Many of you who know about computers know that S.M.A.R.T. is a mechanism that allows people to diagnose hard drives that are on their way out, frequently before catastrophic data loss.
Here’s my question: Does it work?
Are there statistics showing that the technology, in at least one type of drive, has a given level of false positives? False negatives? Failures to predict?
Ok, that first sentence was contradictory the way it was written. Change “frequently” to “allegedly”.
I’ve seen no statistics one way or the other that state that SMART has a certain % of being reliable, but in my experience if you start getting SMART errors your hard drive is in its death throws and its time to backup data now if you don’t have a recent backup. I’ve seen SMART errors on three drives in all of my years of computer support (I do this for a living) and all three have died a few minutes to a few hours later.
Wow? Only 3.
I work in a big, big company’s tech support center. I’ve probably seen 3… in the past 2 weeks. Then again I’m no longer a first-line technician, I’m now the guy who gets called when things have already gone hinky.
I saw a vendor selling a product that had a 5% rate of hard disk failure, according to SMART and a piece of remote monitoring software that was being sold. That would have been OK. Except that the 5% number was on… a monthly basis.
Turned out the firmware on this product was performing some operations that reduced the MTBF of the drives rather substantiallly. A few flash upgrades later, and everything was right as rain.
Most of the systems I worked on were for medical workstations and used SCSI hard drives. For the IDE systems we had I used Quantum, then Maxtor drives at the time. SMART errors were few and few between, but it could be because the systems ran 24x7 so the drives never really needed to park and spin back up.
Sorry, I don’t have the kind of data you require. I had very few hard drive failures, but this was before several manufacturers dropped the warranties to 1 year on ATA drives which would lead me to believe that they expected more failures as well, or they would had left the warranties alone.
==quote===
mbacko1 Most of the systems I worked on were for medical workstations and used SCSI hard drives. For the IDE systems we had I used Quantum, then Maxtor drives at the time. SMART errors were few and few between, but it could be because the systems ran 24x7 so the drives never really needed to park and spin back up.
Sorry, I don’t have the kind of data you require. I had very few hard drive failures, but this was before several manufacturers dropped the warranties to 1 year on ATA drives which would lead me to believe that they expected more failures as well, or they would had left the warranties alone.
==end quote===
My application is 24x7 use of ATA hard drive systems with a few writes per minute. On some units, 30 writes a minute… on some, probably 5 writes per minute.
Occasional database queries against 4 GB databases housed on systems with 256 MB of RAM.
I have had very good success with SMART drives. I think they tend far toward false negatives, but when dealing with data it’s usually better to error on the side of caution.
The only major problem I’ve had with a SMART drive was one instance where I was trying to recover data from a failing laptop hard drive. Rule #1 when dealing with a SMART drive is to find a utility that will turn the SMART off while recovering data. I was using such a utility but was unable to recover data from the bad sectors. Turns out I needed a newer version of the utility, because the version I was using didn’t really know how to turn off SMART on that particular drive. The SMART was preventing the utility from accessing the bad sectors, but it did this “invisibly” to the program, so the program didn’t know the data it was getting wasn’t actually from the location it was accessing. The new version of the utility worked great, and I was able to recover the files I was interested in. The transparency of SMART can be a pain, but without adequate software support there’s no way for the hard drive to communicate such things.
SMART is nowhere near perfect, but it’s better than nothing and basically free. Do they even make hard drives without SMART anymore?
Just my $.02. If I had my way we’d all be using ECC RAM too…