A Mac book I'm reading recommends bailing on DVDs and just using hardrives for backup. The logic is that maybe down the road, macs wont be able to even hook up to optical drives.

I have a lot of redundant backups now with hardrives onsite and offsite, and even in the cloud as well, but still I think of DVDs and kind of a real tangible layer of backup, like a photographer having real film in addition to digital copies. Any opinions on this?

Would be nice to blow off burning those DVDs!

Perhaps you could compromise. Don't back up your complete hard drive to DVD's. That is a real nuisance; I used to do it and can related. Maybe back up only a few items to DVD's, things like selected photos or critical documents. Much less work, far fewer DVD's.

Thanks Ira, no, I was not thinking of that. A lot of the hardrive has been burned to DVD incrementally over the years.

I'm a photographer shooting digitally now. I'm have everything on hardrives and now just started also with Crashplan -- first cloud based backup started yesterday. Would you also burn DVD as well..?

No I would not use DVD's in your situation. But as a parallel alternative you could keep your camera's SDHC cards (or whatever type they may be) with the original shots; don't erase them. This may be overkill whose only purpose is to give you peace of mind with something in hand.

Ira, that's funny because blank DVDs are a fraction the price of SD cards..

Maybe you should look into M-DISCs.

Comparing item for item yes DVDs are much cheaper, but if you compare them on a cost per GigaByte basis not so much.

Thanks but SD are basically hardrives, and I have hard drives covered... Maybe I'm so used to burning those DVDs is hard to get my head around to not keep doing that as back up.. but I leaning now to blowing it off.

Never heard of Mdisc. Do like the idea of 1000 years archival. At this point though I'd want those disc to be in Terrabytes now that I'm going all hardrive.

Nobody's addressed your original suggestion/concern that some day in the future Macs may not be able to communicate with optical drives.

I imagine it's a possibility, but I think that if it ever comes to pass it'll be with more than enough lead time for everybody with optical media backups to switch to something else.

In short, I wouldn't worry about it. (I can't imagine it happening any time soon.)

and a fraction of the capacity, and a fraction of the speed, and a fraction of the compactness...

I see precious little examples of practical, sensible use of optical media for backup nowadays.

I can see at least a bit: a good DVD will likely retain data longer than an SD card.

Flash memory stores information on floating gate transistors. Those pesky laws of quantum mechanics mean those gates will slowly lose charge over time. How long? Hard to say, but certainly decades at most. A good DVD stored in a cool, dark place should remain readable much longer.

Tacit, thanks forget about flash cards. Buy tons of SD cards, and use them only once?

The question Tacit is: keep doing DVD backups in light of the drop in price of hardrives? And online backup.

I just bought 4 new hardrives, all 5TB except the time machine is 8TB, and I signed up with Crash Plan too.

For even faster access and better reliability take those 4 new 5TB HDs and put them in an external enclosure, and use SoftRAID to create a RAID 5 array. Then eve if one of the drives fails you simply replace the failed HD and SoftRAID will automatically rebuild everything that was stored on the failed drive with no interruption in service.

Thanks Joe, I've been reading about Raid for 15 years and still don't get it and that did not change with you post!

I have 1st generation data on one drive called MAIN 5TB. (this is partitioned for the Mac OS 500GB) and 4.5 data.

Clone drive for that, gets cloned 1x day is called: BACKUP DATA (also is 5tb)

TM is on TIME MACHINE 8TB

And I have a 5TB off site.

Why would a Raid be better or beneficial?

Whether you get it or not, RAID 5 works. Suffice it to say I have a RAID 5 array as a Time Machine drive. It is very fast and very reliable. The "trick" is the data is distributed across the various disks with a reliability algorithm that enables the contents of any one of the disks to be recreated from the contents of the remaining disks. Speed is gained because reading and writing is buffered across multiple disks. I actually tested my array by removing and replacing a disk while the array was running. During the recovery it was slower and it did take a while to recover, but recover it did. It won't replace your off site storage but it is far less complicated and labor intensive than your system, my backups occur every hour, and I can recover a given file or even the entire system from any single month, week, day, hour since the backup set was created.And yes I have had to go back a few hours in history to recover the entire system and it worked perfectly.

Thanks JOe, that is still completely over my head. Do you have a photo and/ or link to the set up to buy? I'll keep it in mind for next round - in 5 years. I just bought all these.

What is complicated on my system? I just have three externals sitting behind the computer. SD backs up one, once a day (which has sufficed), and the 3rd is time machine.

Speed seems ok, especially now with USB 3.

Other World Computing has a number of RAID enclosures. I have this one and installed four factory rebuilt 7,200 RPM HGST 1 TB disk drives for a net 3 TB of storage (the reduced capacity is the result of the data that permits rebuilding any one of the four drives should it fail. The whole thing mounts as a single volume. It is virtually silent and just sits there and does its thing with zero effort on my part.

The 1 TB drives are the largest capacity 7,200 RPM 2.5" drives available. For my purposes I could have gone with higher capacity 5,400 RPM drives and any speed loss would only have been noticeable when running benchmark programs but I didn't really need the extra capacity. I figure that at some point, when prices get low enough, I will replace the spinning rust drives with potentially higher capacity SSDs in the same enclosure and that should be faster than virtually any external buss can deliver and maybe even faster than the internal data buss in my Mac mini.

FULL DISCLOSURE I have no pecuniary relationship with Other World Computing other than that of a user of their products.

The nice thing about RAID is it protects your data. Here's how it works:

Suppose you write a letter to your mom. It starts with "Dear mum, life is fine." You save it as a file on your hard drive. That means somewhere on your hard drive is stored a series of characters: D,e,a,r, ,m,u,m...

Now the hard rive fails, for whatever reason. Maybe a bit of falling debris from the space station lands on it, whatever. That hard drive is destroyed; everything is gone. So far, so simple.

Now, a RAID array is a whole bunch of hard drives connected to a special controller so the computer thinks they are just one hard drive.

Let's say you have 4 hard drives in your RAID array. You save a letter to your mum. There are now data stored on your hard drives, but in a very special way:

Hard drive #1 has "D". Hard drive #2 has "e". Hard drive #3 has "a". Hard drive #1 has "r." Hard drive #2 has " ". Hard drive #3 has "M", hard drive #2 has "u", hard drive #3 has "m", hard drive #1 has "," hard drive #2 has " ", hard drive #3 has "l", hard drive #2 has "i", hard drive #3 has "f", hard drive #1 has "e"... Basically, the file is broken up into pieces and little tiny chunks of it are stored on three of the hard drives.

What is stored on the fourth hard drive? Ah, that's the magic.

The fourth hard drive stores the result of mathematical operations on the first three hard drives. So you end up with your letter splintered up in small pieces across three hard drives, and special codes on the fourth hard drive that describe mathematically what is on the first three.

Why do this?

Say that one of the four hard drives--we'll call it #2--dies.

Now the RAID array automatically reverses the mathematical equations that were used to create the data on hard drive #4. So you go to read the letter to your mum, and here's what happens:

It reads the file from the hard drives. But hard drive #2 has been obliterated, so the file looks like this:

D?ar?Mu?, ?if? i? f?ne

Pretty scrambled, right? Every third letter is missing! The file was broken up across three hard drives and now one of them is gone.

So the RAID controller, which is actually a small computer, looks at the information on hard drive #4. It says "I have the letters D and a, but the letter in between is gone because that hard drive is gone. Based on the letters D and a and the mathematical result that was stored on you, what is the missing letter?" And it calculates the answer: It must have been the letter e.

Then it looks at the next three letters. Based on the fact that the next letters are the letter r, then a missing letter, and then the letter M, what must the missing character be? And it does the math and calculates that the missing letter must be a space.

It does this over and over again and it comes up with D,e,a,r, ,M,u,m, ,l,i,fe, ,i,s, ,f,i,n,e. It uses the special codes written on hard drive #4 to rebuild what was lost when hard drive #2 was destroyed.

Tacit, that is a good explanation, did not know any of that.
But, I would be worried that the controller could be screwed up and now I don't have even one good copy! It could screw up everything.

AT least with what I have now:
1 main new HD
2nd backup new HD sitting next to it
3rd backup HD, off site, though I only rotate that every few months
Time machine on a new HD
and I'm starting Crashplan -- which will be another layer in the cloud.

Do I need or want Raid? all of these hardrives can be replaced fairly quickly by a call online. And there are all 100% good until toast. No controllers.

Thanks for that....it solved a mystery for me.

Thanks for the nice, clear explanation of how a RAID array works.

Echoing kevs's fear, though, what happens if the controller fails?

And in the unlikely event that the controller and another drive fail simultaneously...?

The bad thing about a conventional hard drive is that the controller is actually part of the drive (it's on the little circuit board that's mounted to the drive). If the controller on a normal hard drive fails, the drive might be toast. Sometimes it's possible to pry the circuit board off another, identical-model drive and put it on the one that's failed; sometimes it's not.

With a RAID array, the controller is a separate unit. Most RAID arrays are a box that contains the controller and also a bunch of slots for hard drives.

That means if the controller fails, you just send it off to be fixed, or you simply replace it. Each component is designed to be replaceable in a catastrophe, so if a drive fails you pull it out and put a new one in its place (and the controller automatically formats it and copies the data onto it). If the controller fails, you slide it out and replace it. If the power supply fails, you slide it out and replace it. Some RAID arrays even have spare controllers and spare power supplies.

Protecting data is the name of the game with RAID. They're designed so that failure of one single part won't destroy your data.

More good info thanks Tacit.
As Artie ask as well:

What is the controller fails, then you may not even have one good drive? Keeping things separate usually only one drive dies and you quickly replace it.

That controller on the Raid, sounds kind of scary in that it's decoding and then coding back garbly gook, and it that goes screwy the whole house of cards could fall down?

Thanks for the clarification.

So the data recreation process can be bi-directional...drives to controller and vise versa, and a backup controller is icing for critical applications that can't afford to risk simultaneous drive failures, and for worriers.

That is a significant problem. If the controller fails, it can be challenging to access your data. Also some controllers are more robust than others when dealing with hardware failures.

I worked with one customer that had a "network storage" unit, that mounted an SMB share on their LAN to use. The drive decided not to mount up on the network anymore. So I pulled out the drive, expecting to just copy off the data, but noooooooo, they had to use some custom in-house file system. Hours of looking later, I found there was actually a company whose only business was recovering data from this specific model of drive, for a high price of course. Further searching found several users had reported that if it finds corruption, it may spend the next several HOURS fixing it before mounting. As in, maybe a day or two. OK, so we let it sit plugged in over the weekend, as it provided absolutely no indication of status or progress while it was in failure mode. It suddenly popped onto the network mid Tuesday. "Copy all your data off that drive, right NOW" We promptly replaced with with a Drobo. Drobo isn't any better for that, but at least if the controller dies you can get a replacement from them as they're still produced.

Then there was someone that brought in a Mac Pro with two internal drives in a software raid that refused to mount. It was a stripe, meaning two drive interleaved together, with no redundancy, so two 1tb drives looked like one single 2tb drive and accessed faster. (read and write) But it had a corrupt directory and disk utility would neither repair nor mount it. NO disk tools I tried could touch it until it was mounted. Can't mount it until it's repaired, can't repair it until it's mounted, Catch-22. Fixing that required my researching Apple's striping methods, and writing a custom program to re-integrate the stripes from both drives onto a single 2tb drive. One drive had IO errors, so I had to plow through those, but it wasn't a big deal. (that was what was preventing Disk Utility from repairing it, the io error was inside a part of the directory) Once on a single drive, (18 hours later) Disk Warrior quickly repaired the problem.

So raids make me a little twitchy. Never consider a raid to be a backup, that's not it's job. Raids have three jobs: (1) protect uptime by allowing you to swap out a failed drive and keep working while it rebuilds instead of being down while restoring from backup, (2) improves read and/or write speed by spreading access over several drives concurrently, or (3) combining several smaller drives to make one larger single logical volume. Nowhere there does it say "backup". Always back up your raids.

Here is an extremely good recount of a large critical raid failure from initial problems all the way through resolution, I encourage anyone interested in raids to watch it:

https://www.youtube.com/watch?v=gSrnXgAmK8k

Tacit's description of RAID 5 is the most easily understood I have seen.

But in the case of my system there is no special hardware controller, rather the control is handled in software. Theoretically a hardware controller should be faster than software control and at one time that was easily demonstrable but with faster computers that speed advantage is getting pretty slim. Today the limiting factor is more likely to be the speed of the connecting buss and not hardware vs software controllers. SoftRAID in fact contends their software control can rebuild a damaged volume in a RAID 5 array much faster than competing hardware controllers — I hope I never have to find that out for myself

RAID software has been built into OS X for some time now but it supports only RAID 0 (Striped) and RAID 1 (Mirrored). SoftRAID supports RAID 0, 1, 4, 5, and 1+0. (Go here for a description of the various RAID levels and their relative performance and features.)

The purpose of a RAID array is to be resilient in the face of failure. If any part of a hard drive fails--heads, platters, controller, anything--the information on it is gone. A RAID array can tolerate a failure and keep going.

Most RAID arrays these days use pretty standardized parts. They're not really a house of cards; they're quite robust. Years ago, I installed an array for a client of mine, an advertising agency that was at any given time working with many gigabytes of files for a catalog they produced. The array had nine drive slots, two controllers, and two power supplies. It was designed so that if a controller, a power supply, or any two hard drives failed, you would still have all your data. (If one hard drive failed, it would just keep on going; if two hard drives failed, it would go into an emergency mode where you could still read your data but you could not write new data until you replaced the failed drives.)

As Virtual1 says, they're not backups; they're designed to make your data accessible even if something fails. You still aren't protected from doing things like accidentally deleting a file. Even with an array, you still need backups. Think of an array as a hard drive that is very, very difficult to kill.

So Tacit, I was wondering what your take is on the video I posted? Have you had a chance to watch it?

I just watched it, and WOW!!!

No backup scheme for irreplaceable data?

Nuff said!

(I wonder how much it cost them?)

I've only watched part of it. I'm back in the cabin writing another book[1] at the moment, and the cabin has very slow Internet. It's many miles from the closest town, and we have DSL that averages between 55 and 60 kbits/sec; roughly as fast as a dial-up modem. So YouTube is tough.

From what I saw, they are poster children for How Not To Protect Data and How Not To Handle a Failed Array, though.

[1] Actually, two books, one fiction and one nonfiction.