Are USB hard drives really reliable enough to backup to?
Solution 1:
"...part or all of their backup solution."
USB HDDs for part of the solution is okay. For all of the solution, it's not a good idea.
When you plug in your USB HDD, your entire backup medium is now online and writable, on the same machine which has your primary data.
In the event of:
- a glitch in your backup software
- an electrical failure
- a virus
- human error
- fire or flood
You could lose all your data, even your backups.
One poor-man's solution would be to use two HDDs, keeping one off-site and off line (e.g., at a friend's house).
The disks are reliable. Some USB electronics are flaky, so test your backups.
Solution 2:
I have been monitoring backups to USB drives for three years after I noticed too many failures with client's drives. Monitoring the event log on the computer being backed up has been revealing. Even with backups that are successful, write errors are recorded in the event log. Subsequent disk checks of the USB drives revealed either bitmap errors or MBR errors. Since any disk error will eventually lead to data corruption, it is a disaster waiting to happen. Even though the backup may be "successful" according to the backup program, write errors or disk errors count as a failure as you are playing russian roulette with your data. As Journeyman said, the point of backing up is to not have failures. Backing up a 99.999% reliable device with a 90% reliability device does not make sense. So I have opted for backup to NAS onsite and offsite backup through the internet.
The reasons for a high USB failure rate are actually quite simple, but not always possible to avoid.
Failure to remove drive through "Safely Remove Hardware" icon.
A full computer backup through software such as "Shadow Protect" (my favourite) push through large amounts of data at high speed. While most USB drives are sold with high data rates, the reality is that sustained high data rates create heat in the USB electronics. And since most USB drives do not have fan cooling, the electronics go over temp and stop until they cool down. Some drives are better but they all do it.
Again because of high data rates and no fan cooling the actual hard drives quickly go over the rated internal temperature, the temperature of which varies from drive to drive. And the higher the temperature the quicker the rate of temp rise. Hard drive life is shortened but an instant killer is immediately unplugging the drive while it is still hot.
The way people handle or abuse USB drives causes pre-mature failure.
The failure of some USB drives to properly dismount, so the client unplugs it anyway.
I have had to recover data from countless USB drives for clients, and the failure rate is too high. Several times, when client computers bit the dust, a restore from a backup USB drive also failed.
As per Ligos, taking the drive out of the USB case will help but this is not a commercial solution. Some drives still get piping hot even out of the case, and forced air cooling is necessary to get the drive under max rated temperature. I use HDSentinel to monitor temperature and also for general drive health checks.
There use to be small Verbatim portable drives that had a fan and connected by Ethernet but they are no longer available. For the moment it is backup to NAS and online backup for critical data.
Solution 3:
I've had 2 USB HDD's fail on me in the past. Both due to HDD failure (the USB enclosure is still working; verified by plugging the HDD into another computer).
100% of my failures have been due to accidently dropping the drive, having it fall off a table, having the cable wrenched out at unexpected times, etc. Basically, my own negligence and stupidity.
I fixed this problem by taking the HDD out of the USB enclosure and connecting it via the internal SATA interface in my home media PC. There have been zero failures over an equal period of time since I took me out of the equation!
Remote backups are done via a remote backup service over the Internet.
Solution 4:
The point of backing up is not to not have failures. Its to not have a failure when you have another failure going on at the same time, and the thing that failed is needed to fix something else thats failed.
Nothing is failure proof, you're playing with failure curves. Internal drives fail, external drives fail, every so often entire BUILDINGS AND INFRASTRUCTURES fail.
So, yes they are. Just don't entirely trust the sneaky buggers and have enough backup backups to sleep well at night ;p