Have, and properly use, a UPS.
by genewitch - 12/22/12 6:41 PM
I had this same suspicion. I have drives that are from the late 90s that still copy data. Sometime around 2005 i noticed all my newer drives failing at an insane rate. This prompted me to buy server drives. This year, someone finally knocked it into my head... brownouts and power outages are hard drive killers. They always have been risky (even to 90s hdd), but new drives spin faster, hold thousands of times more data. The accuracy of the heads is paramount. One too many times of it striking the back because of power failure and even without a head crash there's so many read errors the drive is "dead"
So here's my handy tips:
Buy a Kill-A-Watt or other reliable such device. Do it, right now, regardless of what else you do. They're very cool and using one for a few weeks will tell you exactly how much KVA you need in a UPS. It will save you money and a headache knowing all of this in advance, and you can be 100% sure your stuff will be protected.
Buy a UPS that can run your entire setup for at least 30 minutes. Don't include printers, speakers, or anything that isn't necessary to your computer remaining on. If you buy a really good UPS, you can leave your low power monitor, network switches, and ISP gateway device plugged into battery backup. I can't say one way or the other if you need pure sine or not. Having a working UPS on your systems is enough.
Let your hard drives sleep. All 12 of my local drives sleep within 10 minutes. I'm slowly adding SSDs to all systems for caching (including filesystem), and readyboost for windows. Readyboost especially seems to allow drives to more quickly sleep throughout a normal session. If the drives are asleep and there's a power failure, there's no chance of mechanical damage. Zero. Might fry a board but you won't lose data.
On NTFS or FAT32 mechanical drives... Let windows defrag it whenever it wants to. Especially your boot drive! If you're paranoid, use a FLOSS or shareware version like defraggler. But schedule it and let it run. Remember the faster spinning and imperative accuracy? Thrashing the heads back and forth over a drive every time it accesses something will probably fatigue the parts much faster. The MTBF is only an average, little old ladies who only use their PC to play slingo on the weekends will generally have drives that last much longer than average, and people who have 64+ gigs of swap and hibfiles on their drives who never defrag will have drives that last a much shorter time than average.
I've had 4 power failures since implementing this on my own hardware. Two brownouts (where the voltage drops for an indeterminate amount of time) as well. Not a single drive has died or even caused me concern. Yes, i have to wait an extra 4 seconds for that recorded TV show i have on the media server, but at least i still have the recorded tv show.
Thanks for reading!