This month Google released their compiled data on hard drive performance from their datacenters around the world. When you’re Google, this type of study is far beyond the anecdotal reports that can be found on gear forums like “I picked up these piece of $&@# Drives from (xyz low-end drive manufacturer) and one already crapped out!”
Instead, Google basically posted “We’ve got these datacenters all over the world with over 100,000 cheap SATA and PATA drives between them and a bunch crapped out!”, oh yeah and here is some great insight into the circumstances under which this occurred. Over 5 years, they recorded every failure and many variables and then repurposed it into an independent report unprecedented in size.
Here are some of Google’s hard drive findings:
1) Temperature surprisingly didn’t have much influence on failure rates.
Warm or cold, Google didn’t report much statistical variance here.
2) Older Drives fail more often but drive use didn’t factor into it as much.
We would be concerned about this as a used IT reseller, except we don’t see many hard drives lately anyhow due to privacy concerns. When we rebuild systems these days we seem to primarily use new drives from the manufacturer to complete configurations.
3) Some cheap drives are better than other cheap drives
At one point Google mentioned that certain manufacturer’s drives definitely failed more often than others. Unfortunately, they didn’t mention which manufacturer’s hard drives failed though and it is probably not in Google’s best interests to share that information. If that ever leaks out, it could be really damaging to a manufacturer like IBM, Seagate or Maxtor.
4) SMART is pretty smart (most of the time)
S.M.A.R.T. (Self Monitoring, Analysis and Reporting Technology) scans that show a drive problem are very likely to fail within 60 days. Google determined that after receiving a SMART error, drives are 39 times more likely to fail than a drive that isn’t reporting one. Unfortunately, many other drives will also fail that never show a SMART warning (36% of SATA disks that failed were never detected by SMART).
A Slashdot contributor questioned whether or not Google’s study was useful, but I definitely think it is. As one commenter “spineboy” stated “To me it’s useful – if I get a SMART warning, then I’m definitely backing up my drive and will replace it before it croaks.”