[NLRS] Hard Drive Backup & Redundancy (was: WA2VOI DISASTER)

[John P. Toscano]

Jason Godfrey wrote:

> OK, I used a bad choice of words. I don't consider RAID a "problem".
> (In fact, I've been trying to convince a relative to get some business
> critical stuff on to RAID.) But I've encountered people that assume
> that because they have RAID they are safe, which isn't true. I've seen
> as many computer have to be installed from scratch due to software
> issues as from lost drives. RAID won't help you with that, backups
> will.
> 
> RAID is very useful tool. It is not the end all solution to that which
> threatens your data (in which I include applications.)

RAID can be protective of your data, or can increase the risk of data 
loss, depending on which "level" of RAID the disk array is configured for.

In brief, RAID 0 (striping) is a performance-oriented solution, but a 
2-drive RAID 0 array is twice as likely to fail as a single drive, 
because failure of either drive renders everything unusable and 
unrecoverable on all the drives.  And a 4-drive RAID 0 array has 4 times 
the risk of failure, and so on.  Just for the sake of simple numbers, if 
the failure risk of one drive is 1:1000, the risk of a 2-drive RAID 0 
array is 2:1000, etc.

RAID 1 (mirroring) is a reliability-oriented solution. Each primary 
drive has an exact mirror image on a secondary drive. If either drive 
fails, the other drive still has all the data, so no data are lost 
unless both drives fail. If the risk of failure of any drive is 1:1000, 
then the risk of failure of a RAID 1 array is (1:1000) x (1:1000) = 
1:1,000,000.  The main downside is that your net storage capacity is 
half the size of the sum of the individual drive capacities, e.g. two 
100 Gb drives add up to 100 Gb of storage.

RAID 5 (striped set with distributed parity) requires a minimum of 3 
drives, but I would say that a 4-drive array is probably more common. 
This can yield most of the performance of RAID 0 with most of the 
reliability of RAID 1. In a RAID 5 array, if one drive fails (or is 
pulled out), the distributed parity allows the remaining drives to 
compute the missing bits, and provide all of your data transparently. 
You then pop in a fresh working drive in place of the failed one, and 
the array re-builds itself.  For a 4-drive array of drives with a 
failure risk of 1:1000, the net chances of data loss would be (1:1000) x 
(3:1000) or 3:1,000,000.  This is because the first drive failure 
(1:1000) causes no data loss, but if any of the remaining 3 drives fails 
(3:1000) before the array is re-built, then your data are gone. An array 
of four 100 Gb drives has 300 Gb of user capacity, with the other 100 Gb 
devoted to the parity information. Surrendering 25% of your capacity for 
security is a nuisance, but one I happily pay. And it beats the 50% 
capacity loss of RAID 1.

There are many other RAID "levels", some standard and some non-standard. 
  You can read more about it here:
   http://en.wikipedia.org/wiki/RAID

For my money, RAID 5 is the way to go.

Also, FWIW, it is true that a lightning hit is probably going to wipe 
out all the data that aren't stored completely offline or in a remote 
location. But in my 29 years or so of personal computing, I have never 
lost a drive to lightning (even though for the last decade or so I have 
had a lot of aluminum pointed at the sky with coaxial cable bringing the 
captured energy right into my house), but I've lost at least a dozen 
hard drives to drive failures of various other sorts, such as head 
crashes, bearing seizure, etc.  Now, I'm not discounting the possibility 
of lightning-induced data loss, as I personally know several hams whose 
homes have suffered direct or near-direct hits and who lost virtually 
everything electronic.  I'm just saying that drive failure from other 
causes is much more likely.

So, for my money, a RAID 5 array is a useful solution, but anything 
critical still gets backed up elsewhere. In the old days, that was 
floppy disks, but my first hard drive was a 5 megabyte model, which 
became a 3.75 Mb hard drive when one of the 4 heads crashed. So, backing 
up onto 100 Kb floppies (yes, they really WERE that low in capacity!) 
was doable. Later, as the hard drives became larger, I tried tape 
drives, 90 Mb and 230 Mb Bernoulli drives, 120 Mb LS-120 "Super 
Floppies", 2 Gb Jaz drives, 700 Mb CD-recordable drives, etc. Nowadays, 
most important stuff gets burned onto a 4.7 Gb DVD-recordable disc, or a 
9.4 Gb dual-layer recordable DVD.

If you are contemplating a system with backup drives that are externally 
mounted, keep in mind that a USB (480 Mbps), FireWire (400 Mbps) or 
FireWire 2 (800 Mbps) interface is reasonably fast, but no match for the 
intrinsic speed of SATA (1500 Mbps) or SATA-2 (3000 Mbps).  But if your 
computer doesn't have eSATA connectors, you will have to go with a 
slower interface. And if your PC has a PCI bus, you can't plug in a PCI 
card with eSATA ports that run at full speed, since the PCI bus can't 
keep up. You'd need a PCI Express card to keep up.  For my external RAID 
5 array, I chose one with a Gigabit Ethernet interface. It runs at full 
speed with my desktop PC or my high-end laptop, both of which have 
Gigabit Ethernet built in, but it will also run at 100 Mbps if plugged 
into a computer with 100-Base T Ethernet only.

73 de W0JT