Wednesday, May 24, 2017

How to back up your data at home, Part II - A brief digression on backups

ASM826 emails:
Rule 1: Every mechanical device fails. 
Rule 2: Every hard drive is a mechanical device. 
Rule 3: It's the data you care about.

Redundancy. Don't back up once because two is one and one is none.
A fairly new technology getting more widely deployed is Solid State Disks, storage that is made from memory chips rather than rotating disks.  This helps a bit with Rule #2, but only a bit.  Sure, there's no drive, platter, and spindle anymore, but if your house burns down you have lost the SSD just like you would have lost your traditional hard disk.

Two is one and one is none.

So think about your data, and how many copies you have:

1. You have a copy on your computer.  This is one copy, but remember: one is none.

2. You have one on your backup device, once you've backed it up.  That's two copies, but two is one and one is none, amirite?

3. You need something here.  Long time reader, friend, and biker dude Burt emails about this:
Be very careful with single-disk backup solutions.  If the single drive dies, you're screwed. 
If you're gonna use a locally-connected backup system, at least use a RAID-1 (mirrored) system.  That way, if one of the drives fails, you can recover your data and transfer it to another media before the other drive fails.  (I use 2 NAS systems: one is a 2-drive RAID-1 system, the other is a 4-drive "striped RAID-5" system - and some data resides on both systems.)
RAID stands for Redundant Arrays of Inexpensive Disks, and has been around for years and years.  You won't get one of these for $115 at the Big Blue Box store, but you can get twin disk RAID devices for maybe $300.  Each drive mirrors the other automatically (RAID-1).  The downside is that you only get half the storage space; the upside is that you get item #3 above, which gives you your third backup location.

If you get a twin disk RAID-1 backup drive, you will have three copies: on your computer, on the first disk of the backup drive, and on the second disk of the backup drive.  At this point, a lot of stuff has to go bad for you to lose your data.  That can happen, but now you're talking about catastrophe, like your house burning down.  I'll deal with that tomorrow in Part III (off-site and cloud backup).

Two is one, and one is none.  You want to aim for three.  You also want software that will do this automatically for you.  That will be Part IV.


aczarnowski said...

JWZ's recommendations from 10 years ago (!) continue to serve me well. I also agree with his take on RAID for home, especially non-technical, users.

Backups don't need to be complicated. Heck, copy "My Documents" over to a USB drive on your key chain and you're a mile ahead of most other people. Let a thousand copies bloom.

Rick T said...

Local RAID10 array is copy #1, and at home the automatic offsite backup to SugarSync is copy #2.

I have a batch file on my desktop I fire at the end of every session (or day) to copy new files to a file server.

Two drives in the same RAID box is 1 copy to me, more resilient than a JBOD but still only available if the box is running.

Eck! said...

Experience has me well trained.

Copy 1 is the working system.
Copy 2 is a copy of the working system when first brought up.
copy 3 is a copy of that because...

Then we backup of the day to day current system. Yep, three copies.

In all cases there is a 4th copy and that system is backed up to maybe once a month maybe twice. That system is normally powered down and fully disconnected from everything, the trade word is "air gapped". That system is there for any hack, lightning or other major fail. Its a low performing system that can run the exact same stuff only slower, its only function in life is to copy the mainline system.

Why do that. Nearly 4 decades ago I took a lightning hit in the days
when it was all floppy.. Took over 6 moths to recover the hardware
damage (this is pre PC) and that much longer to rebuild the lost software. About 2 years later it happened again, this time copies,
multiple copies in depth. Recovery time for hardware was still long
but the software was well protected (one set in a fireproof safe).
Recovery then was literally fixing the fried cards of a S100 machine.

That was in the late 70s, since then PDP-8, PDP-11, VAXen, and now PCs PCs, got the same treatment. PCs more so since failure rates
are much higher. Networking has made it all the more important to have a distinct copy that cannot be reached by hackers, fire, flood,
or whatever may happen. Its easier if it s a whole system that has a good shelf life even if slow. Its only needed to copy to the new system.

With memory sticks (SD or USB) reaching 128gb or more its easy to get a handful and fill them with the intent of tucking them away.


thesouthtexaspistolero said...

That can happen, but now you're talking about catastrophe, like your house burning down.

And I'm here to say, houses do burn. Sabra got the electronics out, but my backup drive was lost. Still have my Crashplan offsite backup, but that USB backup drive will be replaced ASAP.

Rick C said...

JWZ's recommendation also includes "stay the hell away from RAID", and he has several good reasons why. Some of them may no longer be relevant.

SSDs don't suffer the same mechanical failures as disk drives, but they DO wear out eventually. Probably won't happen to most people, but it can. Also they're more expensive.

Also, any recommendations on software for Windows? I have a 1TB external drive that I have an (outdated) backup created with Windows Backup, but I'd like to see if there's anything better, and, obviously, I need to do it more frequently.

burt said...


JWZ's recommendation to "stay the hell away from RAID" may have held once upon a time, and I would have agreed back when ST256 drives were all the rage. Hard drive technology has vastly improved since then. I honestly haven't had a hard drive failure in... years. But I swap out hard drives occasionally anyhow (more on that further down).

Today, the only difference between a single drive USB-3 backup system available at Worst Buy or Stumples, and the use of a RAID-1 USB-3 backup system (mirrored) available from the same places, is a few bucks. That's it. They both probably use the same model drives.

I wouldn't put SSDs into a RAID cabinet. Ever. Unless you're willing to pay the price for enterprise-level SSD drives, of course. Prohibitively expensive for a home backup system.

At the price of external hard drive systems, it almost makes sense to buy a new backup system every 6 months or so: copy all of the current backup data to the new drive and store the old drive offsite somewhere. And maybe bring it home to update it occasionally.

If you decide to use proprietary backup software, remember this: unless the software stores your files "in the open", you'll have to use that same software to recover your backup. Yuck. I have no suggestions for Windows other than to drag-and-drop your data to the backup system.

Emily Bates said...

If you are looking for a new hard drive, while your previous hard drive is not able to back up the data even if you follow the above procedure you can buy them on iBhejo.