The layman's guide to data backup - please advise my next purchase!
February 25, 2010 8:13 PM   Subscribe

The layman's guide to data backup - please advise my next purchase! 1. Google-eager but totally confused Newbie. 2. Budget limit $750. 3. Want some kind of system so that if one drive fails, another one will still keep everything. After a few days of research, the pool of information is just getting too big and murky. Please help me clear the fog... More info inside.

• Macbook, no firewire port, 10.5.8, 2Ghz Intel Core Duo
• I want to back up on two drives, so if one fails, I'm still good (and preferably easily — I don't know about this stuff)
• I would like to be able to edit movies directly off of it
• I don't care about shape or design or noise
• Budget ABSOLUTE LIMIT $750

I've heard Drobo is good, but can be slow when, say, watching a movie
Drobo S is apparently quicker, but stupidly expensive
Also, Drobo + 2 hard drives = even more expensive.

It seems that I could get three 2TB hard drives, using one of them like a time machine, and the other two to store the larger stuff (plugging in first one, and then the other, thus duplicating the info manually). But if I do that, should I get something called an "enclosure"?
posted by omnigut to Computers & Internet (39 answers total) 20 users marked this as a favorite
Build a fileserver.

Find the cheapest PC you can get that can fit two 1TB SATA drives and is compatable with opensolaris. You should be able to do that for under $750.
If you can fit a third, small drive in there for the OS, go for it. You can probably find an opensolaris based distro that boots and runs off a USB stick, so the drives in the box are used completely for storage and not OS stuff.

Set up an opensolaris zpool of whatever type suits you.
Access it over the GigE port via NFS from your mac.
posted by TravellingDen at 8:22 PM on February 25, 2010

Just as an aside, you shouldn't be directly editing anything that's on your backup media. Backups are supposed to be just that, a spare copy of live bits that exist on your primary computer.

I would get a Network Attached Storage (NAS) device, such as this one, buy two 1.5TB drives and configure them as a mirror (both drives have the exact same contents, and one will take over if the other fails). Use Time Machine to manage the backups to that device.
posted by fireoyster at 8:34 PM on February 25, 2010

Hi TravellingDen, as a writer it's rare that an English-language sentence appears that is 7/10 nonsense to me. However, I'll try and ask intelligent questions:

1. If I use that method, does that mean I'll have to buy another PC when I run out of space (I'm a movie editor as well)
2. Hahahahahahahahahaha. Um, yeah, okay, I'll assume you're mocking me :) I wonder if I should also add in the cost of hiring someone to teach me what opensolaris, distro, zpool GigE and NFS mean (I found the definitions, which didn't help much).

Layman: You mean I set up an intelligent (PC) harddrive, which my Mac can access through a cable if I use the right stuff?
posted by omnigut at 8:40 PM on February 25, 2010

Seconding fireoyster's recommendation. I've got a 323, you buy two hard drives, set them up as mirrored, and then point Time Machine at that networked drive. If one craps itself, you have a full backup on the other drive.

It's plenty fast enough for watching video, even 1080p streams. (16-20 MB / second read and 12-14MB/sec write over gigE, a touch slower over 100mBit)
posted by defcom1 at 8:42 PM on February 25, 2010

fireoyster, thanks for the tip. question: I've heard people say that RAID isn't a backup. Should I be worried about that, or are they talking about something else?
posted by omnigut at 8:45 PM on February 25, 2010

A RAID enclosure using RAID mirroring plus two hard drives is probably the most cost effective way to achieve the redundancy you want.

Any USB 2.0 solution, including Drobo, is going to be limited by the maximum speed of USB 2.0. I would think this would make editing movies noticeably slower.

Drobo-S is faster because it has eSata, but your Macbook probably doesn't.

Gigabit ethernet is theoretically much faster, but that requires more hardware --- either your own fileserver, like TravellingDen suggests, or NAS. You can find NAS with RAID, which may be your best bet. Although one reviewer for that linked unit claims to only get 100mbit speed, so read the reviews carefully for whatever you buy.

The advantage of Drobo over RAID enclosure/NAS is that the hard disks don't need to be the same size (they must be for RAID) and the set up may be easier.
posted by qxntpqbbbqxl at 8:54 PM on February 25, 2010

2nd question. I've read some of these 323 reviews, and it looks good. So, secondary question: I've now got about $300-$400 to play with (on that link it only costs $179). What drives can I get for that, and are there any other accessories I should consider?
posted by omnigut at 8:57 PM on February 25, 2010

You are right, RAID is not a backup. However, the statement only applies if you are RAIDing your primary data storage and have no other copies of that data. A lot of people say "oh, cool, I'll just set up a RAID 0 mirror on my regular machine and now I have a backup!" That is incorrect. Having a RAID mirror for your second copy of the data is very prudent. Having a RAID mirror for the primary copy is also good (i.e. if one drive dies, you don't even have the data loss from the last time you took a backup and now), but it's a belt-and-suspenders approach.

For drives, I recommend the Western Digital 1.5TB Green (WD15EADS) SATA drives. They are part of the green line, which uses less power and generates less heat. You're not as concerned with performance on a backup solution because the transport to the disks (i.e. network, USB or firewire) is almost certainly slower than the interface on the disks themselves.
posted by fireoyster at 9:04 PM on February 25, 2010

Thanks....wait, I'll copy-and-paste that..."qxntpqbbbqxl." Obviously you were named on the same planet that TravellingDen learned how to talk.

Network Storage - is this something I'd get alongside the 323, or instead of?

Side note, for anyone in general, I live in New York. Are there computer gurus sitting around in Starbucks waiting to explain everything to me for $20?
posted by omnigut at 9:05 PM on February 25, 2010

Correction, RAID 0 is stripe set, RAID 1 is mirror. Just being technically accurate. :)
posted by fireoyster at 9:06 PM on February 25, 2010

A D-Link 323 is a NAS (network-attached storage) device. It's basically an enclosure that takes disks and makes them available over the network. This device is easy to set up. Get it, plug in the drives, and go to the web-based control panel to say "I have two disks and I want them mirrored."

After that, tell Time Machine to use that device as its destination for a backup. From what I know of Macs, Time Machine will do 90% of the work for you once you've told the 323 to use the disks as a mirror. :)
posted by fireoyster at 9:09 PM on February 25, 2010

Oh, I'm being an idiot. NAS is also the 323. Sorry.

So, I need to make a choice between this, this and this, then buy two of those green drives, and then I'm good.

posted by omnigut at 9:10 PM on February 25, 2010

Great. THANKS!
posted by omnigut at 9:10 PM on February 25, 2010

Do what fireoyster says, but consider running the two drives independently instead of as RAID-1 (mirroring). With a mirror, you get one logical backup on two physical drives - one drive dies and you're fine... but if you make a bad change to a file and then run a backup, your backup is also hosed. So on even weeks, you backup to drive A, on odd weeks you backup to drive B; that gives you a narrow window in which to discover your blunders.

You can do the same thing by getting a few 3.5" external disc enclosures, putting a big disc in each, labelling them carefully and backing up to them in round-robin order (different disc each day). When not running the backup, the discs live in a fireproof safe and the one that's not currently being backed-up to does not get touched.

TimeMachine is probably your easiest option, but do some reading up on rsync too.

Apologies for the info-overload, but you also need to consider what sort of faults you need to be resilient to, because your requirements will inform what technology you use. HDD crashes are the most common failure and are most simply fixed by running a RAID inside your main machine, no external backup. One drive fails, just replace it and the RAID will fix itself. But what about fire? Theft? For both of those cases, you really need to think about offsite backup, if only for the most critical data. How you implement offsite backup depends on how much data you need to backup, how quickly you're creating it and how often it changes. Shuffling hard drives around to external sites is a hassle, particularly because they're vibration sensitive. The internet works but is slow; online cloud-storage services (like, while awesome, are expensive for large quantities of data.

If you want to be able to go to arbitrary previous revisions of your work, you want to think about revision control software as well as backups, but that's a whole 'nother question.
posted by polyglot at 9:11 PM on February 25, 2010 [2 favorites]

Sorry fireoyster, thank you for the info. My faith in Metafilter knows no boundaries, and yet still I'm impressed at how people like you can get me to understand shit.

Oh, and TravellingDen, I'm also marking yours as a best answer because I just assume the less I understand of something the better it probably is. Thanks!
posted by omnigut at 9:16 PM on February 25, 2010

oooh, looks like I should have previewed...

If using a NAS where the drives will be running fairly continuously, choose drives that are designed for 24/7 use, not consumer drives. For example, WD labels theirs as "AV" drives and Seagate makes a similar distinction between some of their models.
posted by polyglot at 9:16 PM on February 25, 2010

No, polyglot, I'm actually always right, so I'll never need a revision. Each later revision of my work is simply more right than the earlier ones. As for fire, I'll go down in flames before I let my harddrive die without me. And theft? Meet my gun, hobo.

But thanks for the heads up. I think I can fairly easily separate my files into the "Need for Life" and "Need for Now" and "Would Like but not Essential" groups. I've got a couple of smallish (by modern standards) hard drives that can cope with the life stuff, like my writing, which would, as you suggest, need revision-proofing.

posted by omnigut at 9:24 PM on February 25, 2010

Sorry, polyglot, but are you saying that AV means it is or isn't designed for 24/7 use? :)
posted by omnigut at 9:27 PM on February 25, 2010

Considering a WD15EVDS AV-series drive is only $10 more expensive than the green drive I originally linked, you could go with that. I've only used WD's green drives in my backup stuff because I have two of them, and if one dies the other will stay running long enough for me to slot in a spare (though that hasn't happened yet).

According to Apple's support page, Time Machine supports versioning of files (hence the name). You choose what date of the backup you want to restore from.
posted by fireoyster at 9:30 PM on February 25, 2010

WD AV drives (they cost a little more than the non-AV version) are the more-reliable ones intended for 24/7 usage. I think Seagate brands their reliables ones as "ES" but it's ages since I bought any drives so could be wrong on that.

Don't be too sure about the fire/theft thing. Are you home 24/7?

For the revision control, have a look at things like subversion. No need to use a separate backup medium or mechanism - you just backup your version control repository with all your other data and to the same device; the version control makes sure that all versions are available. Works great with text and stuff but not with large binary files like images or video.

I do a lot of photography, which means creating (but not so much editing) big wads of data - maybe a gig or two from a session, 10GB from a wedding and recently, 40GB from a holiday. What I do is run a RAID-5 in my desktop machine, so that protects me from drive failure, copy data to an external drive every so often and then I burn my data to DVD-R with par2 redundancy and post batches of them to my parents' place (3000km away). Burning discs is a chore and they've become poor value ($/GB) recently, but they can be stuck in the post without fear of damage, which makes them a cost-effective remote-backup system. Posting a 200GB batch of backups costs about $10 in postage and $25 in media, which is way less than it would cost me to transfer that data over the internet.
posted by polyglot at 9:45 PM on February 25, 2010

Since fireoyster points out that TM does versioning, just use that and ignore my reference to subversion.
posted by polyglot at 9:46 PM on February 25, 2010

Just a fair warning that most of the NAS boxes listed here will only offer Windows file sharing (SMB protocol) and call this "Mac-compatibility" — SMB is an incredibly slow option for network file storage for Macs. Ideally, you will want a NAS with native AFP support.
posted by Blazecock Pileon at 10:08 PM on February 25, 2010

I asked a similar question a while back with good success. I'll say, it largely depends on what pain you want to avoid. Building a file server is great, but requires some ongoing effort. NAS is nice, but limited in flexibility. No surprises, life is full of tradeoffs.
posted by heliostatic at 10:37 PM on February 25, 2010

Does anyone know of a NAS with native AFP support? I found this article, but it still didn't make much sense to me.
posted by omnigut at 10:53 PM on February 25, 2010

Just adding to the already excellent advice.

Drobo is horribly reviewed. Netgear ReadyNAS is well reviewed. (just visit Amazon & see for yourself)

Network Attached Storage will not be as fast as firewire or USB because most networks are not as fast as USB or firewire.

If you're wireless, it'll be even slower because wireless is slower.

As Blazecock Pileon says - make sure your solution plays nicely with Macs.
posted by MesoFilter at 10:56 PM on February 25, 2010

Buy multiple external drives and use two or three of them for backups. Keep one of the drives somewhere else [not home] and swap it once a month to keep it current. External media is incredibly cheap so there is no reason to try and make a combination use of a given drive. Do not bother with "enclosures" -- if you're asking this question, just buy a ready to use external drive. They are cheap. Assume that your "work" drive used to edit movies is not backed up and move the final copies to your laptop (which is).


1. RAID is not a backup solution. Please stop suggesting it. If you have a RAID NAS, you still need to back up anything on it, so I don't think any of the "build a fileserver" or "use RAID" responses actually answer your question.

RAID file servers can have simultaneous multi-disk failures. This problem is made worse by the very long rebuild times for most of the NAS products on the market -- there is a huge window for the second failure. This is the result of disk density increasing and there is little that can be done about it.

2. If you are a Mac user, Time Machine is your best single tier option. Just buy a large external drive and plug it in. Cost? $100 for 1.5TB at Fry's on Wednesday of this week. If you life somewhere savage you'll have to buy online.

Time Machine is a perfect solution in that after a failure recovery is absolutely brainles since the OS boot CD can access and restore from TM directly (or use it to migrate to a new machine). The UI is IMHO weird but all interfaces to date for managing multiple revisions are terrible and at least for normal users TM isn't terrible.

You can use said external drive for any purpose in addition to using it for a backup, so the editing movies thing is fine if you can tolerate the speed.

Where Time Machine falls apart is it doesn't [really] support multiple backup disks which are rotated on and off site. Time Machine is also not encrypted which makes keeping one of yur backup drives elsewhere kind of icky.

3. Use something like CrashPlan in addition to Time Machine

Like Time Machine, the + version supports incremental continuous backup. The UI sucks though -- really, it sucks horribly. But the datastore can be set to be encrypted. It optionally supports backing up over the internel, but this is not nearly as important as the fact that it supports multiple backup target disks which you can now rotate sensibly.


1 work drive used for scratch (editing whatever)

1 to 3 backup drives

In the single backup drive case, partition the drive as follows:

20 GB - OS install disk clone (clone Leopard DVD to this partition)

20 GB - OS install disk clone for future use (just leave it blank for now, if/when you upgrade to snow leopard clone it here)

REMAINDER GB - all the rest to be used for time machine

In the two drive case, same scheme, but REMAINDER is used for crashplan datastore.

In the three drive case, have two crashplan disks, encrypted datastore, one of which is kept off site and swapped with the other once a month. Keep them unplugged and powered off and plug'em in overnight every few days or once a week. The other one is used for time machine and your scratch disk, plug it in [wall and laptop] whenever you're at that desk.

Btw, the reason to have the OS DVD clones there is that it vastly, vastly, vastly simplifies your life to be able to very quickly boot the OS if you need to. GB are cheap, the convenience is very high.
posted by rr at 3:34 AM on February 26, 2010

Regarding the AFP question -- readynas supports AFP, but you would probably not want to use it, and definitely not for time machine purposes.
posted by rr at 3:35 AM on February 26, 2010

btw, responsed to a similar thread here:
posted by rr at 3:38 AM on February 26, 2010

I think this talk of NAS is a bit overdone. And any talk of RAID is totally useless. RAID is not what you need.

I'd suggest just go buying 3 big firewire or USB drives of the same kind (or buy 3 sata drives of the same kind and 2 external enclosures). One (a) that you hook up to the computer permanently that is the primary drive for your data. One (b) that you leave hooked up to the computer and set up rsync to copy from (a) to (b) every night. Set up the final drive (c) like (b), but don't hook it up right away. Once a week unplug (b) and plug in (c) in its place - and take (b) somewhere else (work).

Basically you have a drive (a), a backup (b) and an offsite (omg my house burned down) backup (c). You backup from (a) to (b) every night and you rotate (b) and (c) every week. If any drive fails you stop everything and immediately go buy another one.

This is what I do and cannot recommend it enough.

Regarding brands - I strongly recommend staying away from Western Digital. I have had nothing but bad luck with them. Seagate has been pretty solid. My current backup system uses Samsung drives.
posted by ish__ at 3:40 AM on February 26, 2010 [1 favorite]

Posting a 200GB batch of backups costs about $10 in postage and $25 in media, which is way less than it would cost me to transfer that data over the internet.

Another option is JungleDisk, which costs $0.15/GB, so that 200GB transfer would have been $30, and you still have offsite redundancy.
posted by spikeleemajortomdickandharryconnickjrmints at 5:02 AM on February 26, 2010

Hahahahahaha. - Sorry, I've just woken up, and it seems I have a new answer, and a long one!

[...some minutes later]

Okay, I now have coffee. And I'm going to find fast mac-compatible hard drives. Seagate or Samsung.

Thanks everyone. And I'll come back later to post what I bought (and how much it cost).
posted by omnigut at 6:14 AM on February 26, 2010

Regarding brands - I strongly recommend staying away from Western Digital.

Agreed. Avoid Western Digital.

If you value reliability, stick with Seagate Enterprise drives. Enterprise drives cost a bit more but are worth every penny. If you just care about low energy consumption/temperature, explore Samsung but please stay away from Western Digital "Green" drives.

All drives fail. Period. With Seagate you'll usually notice problems before the drive fails. Western Digital drives tend to fail with no warning whatsoever.
posted by stringbean at 6:32 AM on February 26, 2010

Drive quality changes year in, year out. There have not been recent DeathStar (DeskStar) mass failures from any of the major players recently and these are your backup drives.

Don't bother buying an "enterprise" drive. If you're doing 2x rotated backup or 3x as described you are _engaging in appropriate planning for failure_. The extra cost is not worthwhile.
posted by rr at 6:57 AM on February 26, 2010

I'm not an expert but I can tell you what has worked for me very well. I'm a photog and I generate an obscene amount of raw files per shoot. My basic backup setup (once I've done all the editing and am archiving it is to copy it to TWO of these:

I have two of these units, each with two 1.5TB in them. It's setup as RAID1 meaning, the drives mirror each other. Whatever I copy to the unit gets copied onto two physical drives for redundancy. Since I have two units, once a week I copy everything on the first unit to the second unit (which also has two drives, so yea! more redundancy.) That way, I essentially end up with 4 HDs with the exact same data. I keep one unit at the office and the second at the house and rotate them. I do have other archival systems in place for "deep storage" of dead data I won't touch except in case of an extreme emergency but this is the mainly what I work with. The eSata makes it nice and spiffy and even using just USB2 it's not bad.

The units are too expensive and I bought and installed my own drives off Newegg. For all 4 drives and both units I paid maybe 550 total.
posted by damiano99 at 7:32 AM on February 26, 2010

Mozy=time machine, encrypted, multiple redundancy, offsite. $5/month per machine, unlimited size. If you're backing up over a net connection, you really want a FAST one. The first backup will take a LONG time (mine took a week running in background).
posted by lalochezia at 9:16 AM on February 26, 2010

RAID does not protect you from accidental erasures, corruption, etc. It is not a backup scheme. Having N different identical HDs is not as good as having N different versions of your filesystem over time.

IMHO, Mozy is absolutely terrible. Terrible support, terrible system performance impact. The first backup takes forever and before recommending it, try out a file restore of 100GB or so. It is a completely broken system.

Crashplan is far better and allows you to buy your way to a fully seeded backup. Carbonite is also better. Mozy seems OK until you actually need to get the files back in bulk.
posted by rr at 9:51 AM on February 26, 2010 [1 favorite]

The disadvantage of Drobo is that it's dog slow. Neither model actually serves data out at either the maximum speed of the interface (USB 2 or ESATA) nor at the speed the discs would serve data out if they were just in an enclosure. An enclosure, since you asked, is a small, usually plastic container for a drive or drives that you will connect a USB, Firewire or eSATA cord for data and most likely a power cable. Drobo benchmark showing ~20 MB/s. Lengthy discussion showing Drobo S over Firewire isn't that much better. In short, video editing means moving around large files (GBs and GBs) as fast as possible, in addition to some pretty serious CPU crunching. Using a Drobo will slow you down. Why get a drive that can transfer data at 100MB/s only to stick it in a Drobo?

Also, if you put all of your eggs in one basket (a Drobo) and the Drobo fries the drives, you're SOL. You could work around the Drobo speed bottleneck (which you would like to since you want to edit movies off of it and faster is better!) by doing what others have said and set up a small local server, but again, you've got no off-site backups this way. Unless your need to have a single volume/partition that is greater than 2 TB, why introduce the complexity of a Drobo or file server?

What you want, and since your budget can accommodate it, is simple, redundant storage, a portion of which can be rotated offsite. You've got no Firewire, so just buy three USB 2 enclosures and three 2TB hard drives.

The first one should be 7200 RPMs as this will be faster than 5400 RPM drives. Newegg's 2TB, 7200 RPM drives. You can probably find some benchmarks out there on the net. This one will be used to edit videos off of. Second, for your off-site backups, you'll get two more drives, one will be a Time Machine backup of your internal laptop drive and one will be a Carbon Copy Clone of the first external drive.

As for enclosures, there are many available. I've gotten four from OWC's selection of USB and Firewire enclosures. I have three of the miniStack underneath my Mac Mini (which is my apartment AV system) and I have one bus powered OWC Mercury On-The-Go Pro for my off-site backup of my important stuff. (I don't feel it necessary to backup DVDs I've ripped, etc, though maybe I'll change my mind if I've got $150 laying around that I wouldn't rather spend some other way.) I like this enclosure because it doesn't need a power cable (one less thing to carry around when I bring it home) and it's quite small, not to mention it looks pretty neat.
posted by Brian Puccio at 10:24 AM on February 26, 2010 [1 favorite]

Don't try to build your own NAS. Frankly, if you don't know enough about it that you have to ask on mefi, it's probably too complicated for you to get right. IMHO it's better to get an all-in-one solution like the ReadyNAS Duo or a desktop (non NAS) RAID unit.

I have a ReadyNAS Duo, and it's great for the money. It has AFP support out of the box.

Others recommended avoiding Western Digital, I recommend avoiding Seagate for WD. It used to be that Seagate was the brand to get, but they have had a lot of problems in the past few years. The best thing to do is check newegg reviews for the particular drive you are considering buying - if there are a lot of reports of DOA drives, then avoid. The WD "Green" drives are fine - you're getting RAID to mitigate failure of one of the drives anyway.

NAS can theoretically be faster than USB or Firewire if you are going over Gigabit Ethernet, but the speed of the NAS unit is often the limiting factor (in the "budget" price range).

Also, all SATA drives are Mac-compatible, they just need to be formatted for use on the mac. Don't pay extra for a "Mac compatible" drive.
posted by kenliu at 1:51 PM on February 26, 2010

Ok - reread through this thread.

RAID and backup are two different things.

RAID (more specifically, RAID 1, aka mirroring) is not a backup solution. It offers "data protection", meaning that if one of the two drives fails, then you still have a redundant copy of the data on the other drive. This happens automatically and immediately without any additional actions on your part. If one drive in a RAID fails, you can pull out the bad drive and pop in a new one, and the data will be automatically copied over to the new drive.

To add to the confusion, there are many other types of RAID schemes that serve different purposes. For example RAID 0 is used to write a single file across two drives for greater performance, but this scheme is less reliable than a single drive by itself.

Backup means that you have another copy of your data separate from your main working files (your "primary storage"). If you do something dumb like delete a file accidentally or your computer gets wiped out by a virus, then you can go and restore your data from that copy. Backup usually requires some action on your part to periodically copy your data or have some kind of automated process to periodically copy the data from primary storage to the backup.

Ideally you will have both RAID and backup. If you have to choose, choose backup. If your primary storage fails, then you can recover your data from the backup.

The ReadyNAS Duo I mentioned does RAID 1 (mirroring), and you can also connect an external USB drive to back up the contents of the RAID. You can also set it up as a Time Machine drive (or allocate part of the storage for time machine), or you can configure it to back itself up over the network to a PC or even another ReadyNAS. Lastly, you can pay for online storage and set up the ReadyNAS to back itself up online (but it isn't particularly cheap either).

The only problem with the Duo is that it is not particularly fast in terms of transfer speed. It seems that all units in the sub $500 price range have this problem (e.g. Drobo, D-Link 323, etc.) I think this is a technical limitation; it's hard to make them cheap AND fast.
posted by kenliu at 2:16 PM on February 26, 2010 [1 favorite]

« Older Engineering toolbox and the design process   |   CRUNK!!! font Newer »
This thread is closed to new comments.