Best eSATA JBOD?

redlandmover writes "I already have an HP Media Server (upgraded processor, and memory) that has already been upgraded internally to 3.5TB. I'm sure everyone already has their favorite backup solution (RAID, WHS, a billion external hard drives, etc). My question is: what is the best JBOD (Just a Bunch of Drives), eSATA-connected, external hard drive enclosure? (Preferably, at least 4 drives.)"

Comments

I stopped reading the summary

[Pop69]

after the cretin suggested that RAID was some sort of substitute for a backup.

Re:I stopped reading the summary

[lobiusmoop]

RAID 1 + swapping out/rebuilding a mirror disk periodically is a perfectly reasonable backup solution.

Re:I stopped reading the summary

[binarylarry]

Wow, that's definitely some iFail.

Re:I stopped reading the summary

[IrquiM]

Although not an off site solution, I agree

Re:I stopped reading the summary

[Anonymous Coward]

RAID 1 + swapping out/rebuilding a mirror disk periodically is a perfectly reasonable backup solution.

Except that your time to bring the backup/RAID1 mirror into sync with the primary RAID1 disk will be far longer than using something like rsync. Your fileserver will be slower because I/O will be flooded with the RAID sync process instead of the much shorter rsync.

Re:I stopped reading the summary

[sjames]

You do know that a RAID can be used for STORING backups don't you? Making your primary storage a RAID is no substitute for a backup. Adding an offline RAID storage can be a backup.

Re:I stopped reading the summary

[Dan Stephans II]

Until your controller goes berserk and craps all over your disk or your other disk fails in the middle of the rebuild. Or...

Re:I stopped reading the summary

[Anonymous Coward]

What's a backup?

Re:I stopped reading the summary

[interkin3tic]

What's a cretin? Some sort of outdated term for a developmental disorder caused by a lack of iodine here used as a pompous insult while announcing Pop69 was so offended by a mistake in the question that he couldn't be bothered to read the rest of it?

Re:I stopped reading the summary

[drsmithy]

after the cretin suggested that RAID was some sort of substitute for a backup.

RAID combined with a snapshotting system (Time Machine, VSS, ZFS, take your pick) can function as an excellent backup system. Not including off-site, obviously, but more than adequate for the typical home user.

I've never really looked into it, but I assume you can configure WHS to take regular VSS snapshots ?

Re:I stopped reading the summary

[deebeed]

I started laughing when he said WHS, BACKUP and data ... storage oh please ;) WHS CORRUPTED DATA FOR A WHOLE YEAR AND MS KNEW ABOUT IT. Do not trust that thing. PLEASE!

Re:I stopped reading the summary

[radtea]

after the cretin suggested that RAID was some sort of substitute for a backup.

I realize that English may not be your first language, but can you point out what makes you think that anything in summary implies RAID is any sort of substitute for a backup?

He's looking for a system on which to keep a duplicate copy of his primary drives. RAID gives you relatively cheap mass storage. Such a duplicate copy is generally known as a "backup".

RAID can be used as media for backing systems up. When it is used that way, it is not a substitute for a backup, it IS a backup.

He's asking if that's a good idea or not.

It doesn't seem like a question a cretin would ask, to me.

Re:I stopped reading the summary

[obarthelemy]

yeah sure.

Let's say it again: Backups are:
- off-site
- offline
- multiple
- tested

anything else is just some kind of high-availability solution, that does NOT protect against catastrophic failure, fires, viruses...

Re:I stopped reading the summary

[Vancorps]

Talk about being old-fashioned. Sorry but you're wrong and all the disk-to-disk backup manufacturers would like to have a word with you.

In all seriousness I'm sure nobody believe you can't have a RAID off-site that is online running snapshots periodically. This protects you from fire, viruses, are equipment failure and at least in my case, allows for business continuity which is pretty important these days.

Of course I do go one further and backup to a 100TB library but thats largely because I don't want to maintain that much online capacity if I don't have to especially since I already had to purchase it once for my main SAN.

Use modern technology, you'll find it much more friendly. Most modern network storage strategies work out great. ZFS does snapshotting making it easy to deploy on small scales. Windows only? Well that's no problem either since you have Volume shadow copy and DFS based on whatever schedule you would like.

I go one further with DFS/VSC and use NetApp snapshotting on the back-end which mirrors the snapshot to another array at another building. Works out great and the only maintenance is the occasional swap out of hard drives when the RAID controllers preemptively fail the drive because they detect abnormalities that will lead to failure.

Re:I stopped reading the summary

[Vancorps]

Realistically its a question most people should ask themselves because it can make a big difference in the cost of doing your backup and also dramatically impacts the amount of time it takes to get back up and running.

Disaster recovery is very much individual to the company. Our company places a high emphasis on survivability so we naturally have a lot of redundancy online, near-line, or offline with variable policies based on the nature of the content. Financial documents have to stay around longer than temporary internet files for instance.

Re:I stopped reading the summary

[LoRdTAW]

Best method I have of backing up my data is simple. First equip/upgrade a few existing computers with 1TB disks even if you never plan to fill them up. They can be at your parents house, siblings house or work. Copy your really important data like work, projects, photos, music, video (movies, tv shows and p0rn don't count), basically anything that is irreplaceable. Copy that to a 1TB USB disk and copy all the data to the computers you equipped with the backup drives. Now you have your data spread out all over. You can use rsync over the net or via a USB disk to keep things updated between machines. You can even partition the large 1TB disks and make a separate partition for your data so it cannot be tampered with. If a machine fails then from any of the others you can replicate the data.

Sounds like a pain in the ass but I keep copies on my brothers PC and my work PC. Its only about 400GB total so its not even half of the 1TB disk which costs about 75 bucks, small price to pay for peace of mind. I have a big software raid 5 array for personal file serving needs but it is in no way shape or form a backup system. I once had my raid 5 go haywire because of some disk controller problems. After a hardware upgrade I almost lost the array but it came back up and had to rebuild itself. Thankfully it didn't send me into a panic because I had my most important and irreplaceable data backed up.

Re:I stopped reading the summary

[lobiusmoop]

If you are rotating your swapped-out disks rather than continually using new blank ones, then the re-mirroring (if done vaguely intelligently) will only update based on the blocks that have changed since the last time that disk was running live in the array (i.e. an incremental update, which is much faster than re-mirroring from scratch).

Re:I stopped reading the summary

[timeOday]

RAID combined with a snapshotting system (Time Machine, VSS, ZFS, take your pick) can function as an excellent backup system. Not including off-site, obviously, but more than adequate for the typical home user.

I disagree, since a single mistake (e.g. mistakenly reformatting the wrong device node, or physically losing the system while moving house) could still take out the whole kaboodle.

And for something you really care about, an offsite backup is worth it and not difficult. I uploaded my family photos to my ISP-provided online file space. If you want to make sure it stays private, encrypt before uploading.

Re:I stopped reading the summary

[Skuld-Chan]

You'll be crying when you rebuild that raid and two disks fail at the same time (happened to me). No - raid isn't a backup solution.

Re:I stopped reading the summary

[Fulcrum of Evil]

RAID 1 + swapping out/rebuilding a mirror disk periodically is a perfectly reasonable backup solution.

Sure, if you're retarded. I was going to say it was ok for home, but no, that's just stupid. Even a batch job that tars a bunch of directories onto a second HD works better (and no additional hardware either).

Re:I stopped reading the summary

[The Archon V2.0]

after the cretin suggested that RAID was some sort of substitute for a backup.

Of course RAID is a substitute for backup. If you ever delete data, accidentally reformat, lose files to a corrupt file system, get infected with a virus, or have any other disaster of the sort, it's obviously something you did or should have anticipated. Thus, data loss is a sign you're inferior and a sinner and the gods of IT are punishing you. Accept their swift and painful lesson with whatever microscopic shred of decorum exists within that rotting, unused thing you call a brain and try to rise ever so slightly above the unenlightened mire of your life so it doesn't happen again.

Besides, it was probably all porn anyway.

Signed,
The cult of BOFH, flagellation division

Re:I stopped reading the summary

[jabuzz]

With modern large capacity drives RAID 5 is just for crazy loons, you need RAID 6.

I would add that RAID anything on a single "shelf" is also playing with fire. I have personally seen an entire shelf of disks failed as a oner, or in *very* rapid succession. Shelf level redundancy is where it's at.

Remember children if it is not offline it is not backed up.

Re:I stopped reading the summary

[master5o1]

I didn't think anyone actually used MacOS X server edition.

Re:I stopped reading the summary

[magarity]

I stopped reading after he uselessly bragged about upgrading the processor and memory. Isn't there a 'lookatme' tag?

Re:I stopped reading the summary

[growse]

So why use RAID? Why not have an offline storage... you know... disk?

Re:I stopped reading the summary

[growse]

Again, why bring RAID into it? A disk combined with a decent snapshotting system can function as an excellent backup system.

RAID use is orthogonal to backup strategy. The two have nothing to do with each other. RAID helps availability, and sometimes performance.

Re:I stopped reading the summary

[tsalmark]

RAID can be used to make two or more disks appear as one larger disk. It's nice when your backup disk is bigger than your primary disk.

Re:I stopped reading the summary

[growse]

You make your point well sir. :)

Re:I stopped reading the summary

[jhol13]

What do you think, does your ISP use RAID or not?

So is RAID good for backups, off-site or on-site?

Re:I stopped reading the summary

[timeOday]

There's nothing wrong with storing a backup on RAID. Just don't assume your primary, working copy is a backup because it's on RAID.

Re:I stopped reading the summary

[sjames]

A single disk is more risky than I would like. Especially since it's offline, it can fail without warning leaving me (unknowingly) without a backup and unable to update my backup until I get a new disk (and hopefully I didn't need any of the archival versions of any of the files)

A RAID is far less likely to suffer that problem. When a disk fails, I have a signal that I should replace enough disks to maintain the RAID even when the remaining old disks fail.

And, as tsalmark said, it's nice when the backup disk is bigger than the primary.

Re:I stopped reading the summary

[LordMyren]

adding more disks is not the solution. the solution is more active and correcting consistency checking. the reason raid 5 and raid 1 rebuilds fail is because data is replicated, but errors are only discovered when something fails, when the shits already hit the fan. zfs, hammer, btrfs, they're all running headlong towards consitency checking because that is not good enough and not pre-emptive enough.

Re:I stopped reading the summary

[Hecatonchires]

I hope that other building is on a different power circuit, as UPS's don't last all that long under SAN load, and out of range of the shockwave of a bomb, or earthquake, or attack targetting your local exchange. (Someone took a chainsaw to the cables in North Sydney a few years ago)

Also, live snapshotting doesn't work when a malicious update eats your database. Those disks are ok, but the data on them is useless. Disk may be cheap, but try keeping a meaningful number of snapshots when one of the ones you support is over a terabyte.

In the end, someone at a higher level than you has to make the call on how much that data is worth, how much rework can be done, and how long the business can be AWOL. It's your job to implement and then support probably.

Re:I stopped reading the summary

[funwithBSD]

I assumed he was asking "What do I use for external drives FOR a backup."

In which case he is a moron for attaching it directly to the system. Or for it being in the same building if that is what he is really after, a backup.

If it is attached to the same system, or in the building/house, what you have is a *copy* of your data, not a backup.
 

Re:I stopped reading the summary

[dna_(c)(tm)(r)]

In the end, someone at a higher level than you has to make the call on ...

And there is the problem. If even the techies have troubles understanding backups, how do you expect managment to understand it?

Re:I stopped reading the summary

[DJProtoss]

I'm not seeing the ambiguity in 'offline' - offline means exactly that, the system is not online (either its turned off or disconnected from the network).

And its real goal is not to protect against such errors (that is what the historical part of a backup like the incremental rsync you described is for). What it is for is protection against a hacker getting into your system and remotely erasing your backups along with your main system (q.f. the recent story re that site of flight sim stuff). Simply saying 'but the target box only exposes ssh/rsync' isn't good enough. All you need is one vulnerability and bang.

You could possibly make an argument in favour of a box which exposes no remote services at all and initiates the rsync itself to a partition explicitly mounted noexec, nosuid; but it is important to realise that that is simply managing the risk down (to ruddy near zero) rather than removing it.

Now for a home setup, its probably the most arguable of all the elements of a backup as to whether you need it. As with all these things, your level of paranoia determines this - I use something similar to what is described above. Yes, I am vulnerable to that type of attack potentially, but that a risk I have judged and made a call on. To think that approach is not at risk is dangerous.

Re:I stopped reading the summary

[dna_(c)(tm)(r)]

If you want to make sure it stays private, encrypt before uploading

How do you backup your key?

Re:I stopped reading the summary

[TheRaven64]

The grandparent obviously doesn't. ZFS is not in OS X Server 10.5, and was silently dropped from the advertised feature list of 10.6 after the WWDC.

That said, ZFS does go a long way towards working as a backup. With RAID-Z plus snapshots, you are safe from drive failure or accidental deletion. You are not, however, safe from attacks that compromise the OS (and are therefore able to write to the disk at the block level), or from things like theft of the NAS or having a power spike frying the controller and all of the drives. You can alleviate these by using zfs send / zfs receive to pipe incremental updates to a remote machine, however this still doesn't protect you from an attacker compromising both machines.

Re:I stopped reading the summary

[drsmithy]

If you are rotating your swapped-out disks rather than continually using new blank ones, then the re-mirroring (if done vaguely intelligently) will only update based on the blocks that have changed since the last time that disk was running live in the array (i.e. an incremental update, which is much faster than re-mirroring from scratch).

I'm not aware of any RAID1 implementation that will do this. Are you ?

Re:I stopped reading the summary

[MistrBlank]

Damnit, there's no to mod for "ignorant" or "misinformation"

Re:I stopped reading the summary

[geminidomino]

That said, ZFS does go a long way towards working as a backup.

Does ZFS make an external copy and carry it across the street?

A backup stored on the same system is not a backup.

Re:I stopped reading the summary

[TheRaven64]

Did you even read the second paragraph of my post beyond the first sentence? The zfs send and zfs receive commands let you sync a snapshot with a copy of that snapshot on a remote system.

Re:I stopped reading the summary

[timeOday]

I don't, I just use a password that I've used other places, and my local backup is not password protected.

Re:I stopped reading the summary

[sjames]

It is a relative term. What sort of "offline" is used depends on the circumstances. For example, keeping the backup server on and ready prevents the dreaded "rm -rf / junk". Moving it to a non-routed secondary LAN helps prevent malicious destruction from outside. Actually powering it down when not in use provides better isolation. Powering off and/or disconnecting it manually (rather than through a PDU with remte control) even more so.

Which is best depends on the nature of the data and operational requirements.

Re:I stopped reading the summary

[geminidomino]

Yes. Unfortunately, I am completely exhausted and didn't realize that my post, without context, only sounded dickish.

It would make more sense if everyone on slashdot magically knew that there was a second fire-safe in the admin office across the street...

Mea culpa.

Re:I stopped reading the summary

[danpritts]

When you say "adding more disks" I assume you mean raid6 vs. raid5.

You're right that consistency checking is required; but it's not good enough for important data.

I did a consistency check last week on my entire RAID. it was good.

I just lost a drive, now i want to rebuild parity. Whoops, I get a read error that crept in since my last consistency check. I am now out my data. RAID6 would have saved me.

If i didn't do a consistency check fairly regularly, i'd probably be hosed even with RAID6 since there would be enough read errors that i'd lose.

Of course, it all depends how important your data is. My MP3s and dvdrips are not so important that a lost file is a catastrophe, so they live on raid-z (like raid5 for redundancy's sake). My business runs on RAID6 and mirrors.

Re:I stopped reading the summary

[PalmKiller]

In related news, a geek was crying all in his cheerios this morning when his raid controller went apeshit during the wee hours and wrote garbage over his only backup.

Re:I stopped reading the summary

[Fweeky]

Linux md RAID-1 (and a few other levels iirc) use a bitmap of changed blocks to reduce the IO required to resync a disk.

ZFS mirroring has similar behaviour, though I think they probably do something fancier with individual records rather than just a coarse bitmap of changes.

HAMMER mirroring certainly does; every write creates a new record ID and replicating systems, be they other disks or other machines entirely, can sync up their filesystem to whatever record they like, keeping around previous modifications depending on how they're configured. Yank a disk for 5 minutes, you get 5 minutes worth of writes for it to catch up on, not a resync of the entire disk.

Re:I stopped reading the summary

[Fweeky]

Backups are ... offline

Shit. Someone tell all those people selling Amazon S3 based backup services.

Re:I stopped reading the summary

[Fweeky]

Print it, put it in a safety deposit box.

Re:I stopped reading the summary

[OriginalSolver]

I really hope you are joking. Here is a talk I did on backups recently for a LUG: http://www.timetraveller.org/talks/backup_talk.pdf Please download and read this. It states explicitly why your statement is false.

Re:I stopped reading the summary

[OriginalSolver]

Network mirroring like this is no backup at all. An error will be carried across by the synchronisation. This is one of several common mistakes made about backups. The system you describe is a variation on "using RAID 1 as a backup system". I am glad to hear you use a tape library as well. I hope the tapes go offsite.

Re:I stopped reading the summary

[OriginalSolver]

Your point is? A lot of backup solutions that are sold are awful and fail to work properly under real world conditions.

Re:I stopped reading the summary

[Fweeky]

You missed the "encrypt it" step.

Re:I stopped reading the summary

[hplus]

ZFS support for OSX: http://alblue.blogspot.com/2008/11/zfs-119-on-mac-os-x.html

Re:I stopped reading the summary

[Vancorps]

Sorry, I didn't use the term mirror. If I did then you would have a point, except that I didn't. Versioning is very important and why you have multiple tiers of storage.

As for the second RAID, it is located offsite along with the tape library so yes, in fact, it is very solid and would take quite a lot of effort to destroy any data.

Re:I stopped reading the summary

[obarthelemy]

fixed it for you:

Someone tell all those people ... BUYING ...

from the Amazon TOS:

"NEITHER WE NOR ANY OF OUR LICENSORS SHALL BE LIABLE TO YOU FOR ANY [...] DAMAGES[...] INCLUDING, WITHOUT LIMITATION, ANY SUCH DAMAGES RESULTING FROM:
(i) THE USE OR THE INABILITY TO USE THE SERVICES;
(ii) THE COST OF PROCUREMENT OF SUBSTITUTE GOODS AND SERVICES;
(iii) UNAUTHORIZED ACCESS TO OR ALTERATION OF YOUR CONTENT.

IN ANY CASE, OUR AGGREGATE LIABILITY UNDER THIS AGREEMENT SHALL BE LIMITED TO THE AMOUNT ACTUALLY PAID BY YOU TO US HEREUNDER FOR THE SERVICES. "

hey, that does inspire confidence... or not.

Not quite what you want...

[Facegarden]

This isn't quite what you want, but I have a $30 6 drive caddy (with 4 drives atm) and a $70 4 port internal SATA card. I just run long SATA cables to it, but it was cheaper than any single-cable solution i found, so that may not be a bad way to go.

One thing I noticed though was that I actually have enough room for all 9 of my hard drives inside my case! I may migrate them in.

And yes, before you say it, that is certainly quite a bit of porn!
-Taylor

Re:Not quite what you want...

[Stickerboy]

>And yes, before you say it, that is certainly quite a bit of porn!

I need some quantification here. Put it in layman's terms - how many Libraries of Congress of porn is that, exactly?

Re:Not quite what you want...

[maino82]

There's a Library of Congress of Porn?!?!?! I take back every bad thing I've ever said about the government and wasteful spending...

Duct tape

[sakdoctor]

Duct tape the drives together, then use software RAID JBOD.
That's what MacGyver would have done.

Re:Duct tape

[roc97007]

Popsicle sticks between the drives, for airflow.

Re:Duct tape

[eldavojohn]

Popsicle sticks between the drives, for airflow.

Yes! This is one of those often overlooked PROTIPs that you have to buy expensive trade magazines to find. I have one here from Onion Technology that says:

Duct tape the drives together in packs of two but be sure to put popsicle sticks between the drives so that the ends of the sticks stick out a little. Use the old fashioned connectors to daisy chain the power to the drives and make sure you have enough SATA ports on your computer. Then, put two or three packs of two around your chest on the outside of a swinging technical vest! You've got a mobile backup device--IN STYLE!

The last part merely involves lithium power source and form factor PC in a backpack on your back add a tiny display and you're set to go to your favorite sporting events and fly in style. Be sure to run up to the arenas to gather attention--you are the 1337 geek of everyone, after all!

Hope this helps.

RAID is no backup solution

[ls671]

Please note that RAID and such are not "backup solutions" ! If your FS get screwed, you loose info.

Think of a backup solution as independent from the media where the info is kept. Then you decide if you want to use RAID, tapes, etc.

My backup solution: incremental backups every half-hour. And full backup once a month.

Now for the media I use to store the backups : RAID mirroring for incremental and hard drives put in a safe at the bank with rotation for full backups. (NO RAID used for full backups).

Duct Tape is a Bad Idea--Use Magnets!

[eldavojohn]

Duct tape the drives together, then use software RAID JBOD. That's what MacGyver would have done.

Duct tape? Oh heavens no! No, here's what I did: I went down to the local thrift store and bought a few big shelf speakers for ten dollars. Then I took them apart and got the really powerful magnets out. Using these, you can attach the drives to the outside of your case. There's one gotcha though--some cases are aluminum which means you have to attach the magnets and drives to your CRT if you have one. This usually just means a longer cable though.

The smart thing about this is that the drives are on the outside of the case so they remain cooler than they would in any enclosure.

If you think a RAID is a backup, you'll be overjoyed with the results of my advice!

Re:Duct Tape is a Bad Idea--Use Magnets!

[jurv!s]

whoosh! please reread the last sentence.

Re:Duct Tape is a Bad Idea--Use Magnets!

[eldavojohn]

You attached strong magnets to your magnetic storage device? Did you intentionally want to erase your data after writing it or are you just an idiot?

Oh well, I haven't had to recover any of the files so far. At least it doesn't ruin my CRT.

Offtopic but how did you get your post to be colored rainbow? I mean, I can't even get regular bgcolor to work on Slashdot ... I know, I know, I must be new here.

Re:Duct Tape is a Bad Idea--Use Magnets!

[Bigjeff5]

I never laugh out loud at work, therefore what I did in reaction to your post was simply an uncontrolled spasm of my diaphragm

Seriously though, that was some frickin funny stuff.

Re:Duct Tape is a Bad Idea--Use Magnets!

[RegularFry]

Wow. That's the first time a comment has actually made me facepalm in a while.

Re:Duct Tape is a Bad Idea--Use Magnets!

[Anonymous Coward]

-1, either too stupid or too clever =(

Re:Duct Tape is a Bad Idea--Use Magnets!

[NotBornYesterday]

You're overthinking the whole thing. The drives themselves have even more powerful magnets already inside, so there is no need to bother with the speakers. Just crack open each drive, and remove one (only one!) of the two neodymium magnets. The drive only needs one ... the other is there for redundancy. Since you are doing RAID somethingorother, you will already have enough redundancy in your storage, and the extra magnets can be used elsewhere.

[/brilliant advice]

I use the Mediasonic ProBox BUT...

[rei_slashdot]

...I can't get the manufacturer to acknowledge or confirm that there is problem when copying between hard drives in the same enclosure. Windows hangs and eventually the Event Logs show "device is not connected" or some sort of issue. Copying between drive and the motherboard's SATA drives works fine but it always hangs/times-out/becomes inaccessible after a random amount of transfer. It's surprisingly well put-together without looking tacky and well-priced but this copying issue between drives inside it is a pain. Transfer between drives inside it seem to work without a hitch using slower USB. http://mediasonicinc.com/store/product_info.php?products_id=150 It syncs with your PC's power so if the PC goes off, the box goes to sleep and wakes up when the PC power is restored.

Wut

[sexconker]

Why do you need an enclosure that does JBOD?
In my opinion you need an enclosure that does 2 things.

Encloses your drives.
Provides power (since current eSata doesn't, LOL).

Let your system handle the JBOD. Everything supports JBOD. Or, you know, just have them as 4 separate drives and be organized, so you can deal with them as raw drives if need be, and so if one goes dead, it'll be a lot easier to get your shit from the others.

I have yet to see a multi-drive enclosure that DOESN'T force it's shitty controller on you, unfortunately.

I would get 4 enclosures and 4 drives.
Stack them on top of each other.
Strap them together with masking tape (less residue than duct tape, provides a good space to write a label, etc.).
Split the output of a single 12v AC adapter (make sure it can put out enough amps) to all 4 inputs.
Run 4 eSata cables to the back of your PC.

Success.
The only issue is splitting the power lead (not hard, but you will need to find the jacks and you'll have to do it yourself) and running 4 eSata cables.

Yes, I'd be willing to do that just to get away from the shitty controllers in external enclosures. Now, if this were SCSI, you could daisy chain the power and the data for the drives.

If only there was a serial-attached version of SCSI.

Re:Wut

[ls671]

Thanks for the tip, I have never used external enclosures with "shitty controllers" but I have been tempted by them. I've only used file/backups servers that I would setup myself with computers running Linux.

Have you actually tried any of these external enclosures with "shitty controllers" ?

Details on problems would be fun to hear about...

Re:Wut

[Ritchie70]

Can't agree with the masking tape. If you don't peel it off pretty quick, you get an awful residue. Duct tape residue will clean off with a little rubbing alcohol.

Re:Wut

[sexconker]

Have never had that issue.
And old duct tape residue (several years) will require significant work to remove.

Re:Wut

[sexconker]

Basically, I don't trust any enclosure to be a hard drive controller as well. Even higher-end boxes. If I want a controller, I'll use my motherboard, or a dedicated storage PC/server. Yes, I have tried many.

Even if they were trustworthy, I would want to avoid them. It's just one more thing to fail, have to update firmware for, track down drivers for, worry about, etc.

This is for your backups, right? Take a minimalistic approach. Just get shit that works. Hard drives work. You're just putting them in a case outside your case with my method.

And I wasn't joking about SAS either:
(Random links for fun. Always fun to look at site other than newegg.)

http://www.shopwiki.com/External+Mini-SAS+to+eSATA+Cable

http://www.pc-pitstop.com/sas_controllers/

http://www.vantecusa.com/front/product/view_detail/354

Slap in some drives and rig up (or find on some Chinese site) a 12v AC adapter with 4 outputs and a few amps.

Re:Wut

[Captain Segfault]

Yes, I am aware that JBOD is also used to refer to simple spanning. It can also refer to drive shelves/enclosures without a RAID controller, which present "just a bunch of disks".

The OP was asking about enclosures, not RAID levels.

Re:Wut

[Ritchie70]

I'm basing it on experience years ago when my car windows were broken out. Used some masking tape and some duct tape to cover the windows with plastic until I could get it fixed.

The duct tape residue came off with a little light solvent. The masking tape put up a real fight.

Of course, getting wet probably had an impact.

Re:Wut

[sexconker]

Informative? WHAT THE FUCK.
You're 100% WRONG, dipshit.
Spanning IS JBOD! JBOD IS spanning!

http://www.yourdictionary.com/telecom/jbod

http://searchstorage.techtarget.com/sDefinition/0,,sid5_gci343350,00.html

http://en.wikipedia.org/wiki/Standard_RAID_levels#Concatenation_.28SPAN.29

Re:Wut

[sexconker]

No, sorry.
JBOD actually in fact means getting a bunch of disks and presenting them as a single logical volume.

It does NOT mean presenting individual disks.
It never has and it never will.

The term JBOD comes from the fact that it is NOT RAID (no redundancy, and unlike RAID 0, which is also NOT RAID, no performance benefits).

Re:Wut

[atamido]

I think you might be looking for something like this:
http://www.amug.org/amug-web/html/amug/reviews/articles/addonics/5x1/

It takes a single eSATA connection and outputs several more. This lets you make your own enclosure or just attach naked drives.

Re:Wut

[sexconker]

That (obviously) requires the host adapter to have port multiplier capability.

I wouldn't trust the onboard sata controllers of most mobos to support that (regardless of what they say on the box). If I had to move the drives to a different system, I might be stuck unable to access my shit.

That, with a compliant sata controller card, would do well to replace the (more esoteric and expensive) SAS card and cable. Plus it's more "put it all in a box"-friendly.

Re:Wut

[Captain Segfault]

I work in the storage industry. If you do too I might be interested in debating this, but otherwise I'll appeal to my authority here.

JBOD, the "RAID" mode, means what you say.

JBOD, the physical object, refers to a device which presents "Just A Bunch Of Disks" rather than doing any sort of RAID. It may, for example, take a bunch of fibre channel disks and put them on a fabric loop, but it would still present the individual disks.

In other words, a JBOD is a generalization of a disk shelf.

Seriously. JBOD is used this way by people who have worked on these things for years.

Re:Wut

[atamido]

I think most recent/decent motherboards have chipsets that will support multipliers fine. What I'm trying to find is a really cheap box with drive trays and a SATA port multiplier. Most of them out there either seem to have a full controller or lots of indicator lights that drive up the price.

Re:Wut

[sexconker]

Seriously, you're wrong.

If you're working with people who use JBOD to refer to individual drives, that's fine. Just know that they're wrong.

You have no evidence to support your claim.

Re:Wut

[sexconker]

You have all the pieces, the problem is just getting it together.

Grab a drive caddy for 4/5 drives from some PC case.
Many of these have tool-less designs.

Grab the power and LED stuff (if you want it) from smaller enclosures (or wire up your own).

Wire the power together.

Hook that board up, and secure it to the side? of the drive caddy.

All you need then is a box and jacks for power and eSata. Maybe a fan.

Re:Wut

[atamido]

Yes, but I'd like to just purchase it as a single nice looking box. The manufacture of such a device seems simple enough that it should be much easier to buy than to build. Meh, I guess other people aren't looking for what I'm looking for.

Re:Wut

[sexconker]

You build it, I'll be your first customer.

Re:Wut

[Anonymous Coward]

Woah woah woah!

          A) Don't blow a gasket there

          B) Even in the wikipeida link you provide, they say "Some RAID controllers use JBOD to refer to configuring drives without RAID features. Each drive shows up separately in the OS. This JBOD is not the same as concatenation."

          C) Again, don't blow a gasket. And you really shouldn't be randomly calling people a dipshit, that kind of makes you a dipshit.

Re:Wut

[Captain Segfault]

When I google JBOD, from the *first page* of results:

http://www.aberdeeninc.com/abcatg/kitjbod-1003.htm

http://opensolaris.org/jive/thread.jspa?threadID=83108&tstart=0

http://blogs.sun.com/brendan/entry/jbod_analytics_example

Re:Wut

[sexconker]

Funny how all MY links were on the first page, and were higher.

Also, your Sun link isn't on the first page.
Also, your Sun link doesn't define JBOD.
Also, your Sun link is arguing against you.
The guy says he got 144 drives in 6 JBODs.

http://blogs.sun.com/brendan/resource/analytics-1/analytics-1-hardware-crop.png

Yeah, that's 24 "1 TB" drives in 6 JBODs.
Nothing in that blog post says what the Sun box is presenting the volumes (externally) as. All this is about is using their fun little GUI to tell that he cabled it wrong (on purpose, to make the blog post).

Your dciginc link also doesn't define JBOD, but it proves itself to be even more wrong than you:
"JBODs use aging RAID5/6 protection schemas to maintain data integrity. Raid 5/6 is becoming more impractical as disk drive sizes grow and rebuild times to recover failed disk drives increase. Further, JBODs provide no or limited self-healing capabilities to detect errors that can occur in disk drives over time and automatically correct them."

JBOD using RAID 5/6? What?

Your aberdeeninc link is for a 4-channel SAS box.
It uses the term JBOD a lot, but it never defines it. It's marketing.

Looks like you couldn't find anything at all to support your claim, yet you posted some arbitrary links to try to save face.

Re:Wut

[Captain Segfault]

Have you even considered that you might be wrong?

You are trying to argue that a term with (at least) two different meanings does not have one of those meanings.

I have provided evidence that people do use the term that way. This isn't some random crackpot, it's on the order of 20% of the google results.

What dciginc link?

The fact that links with the first meaning appear higher on a google search than the other only indicates that the one meaning is more common than the other.

Re:Wut

[sexconker]

You provided no evidence, as I pointed out.

That's right, you didn't have a dciginc link. I was looking at the first page of google results for JBOD and wasn't finding your links.

Your open solaris link is a forum post by some guy.

Dude.
JBOD is a fucking standard. It has a meaning.
I can't go around saying RAID 1 has striping and have it be a valid claim.

JBOD means a bunch of disks, with data spanned across them, presented by the controller to the host as a single volume.

This is what it is. It's a definition.

Re:Wut

[Captain Segfault]

JBOD is a fucking standard.

The closest thing to a standards body I'm aware of for RAID would be SNIA.

http://www.snia.org/education/dictionary/j/

They say:

Originally used to mean a collection of disks without the coordinated control provided by control software; today the term JBOD most often refers to a cabinet of disks whether or not RAID functionality is present. See disk array.

Note that this actually does *not* match my definition, insofar as it includes shelves with RAID -- but it doesn't even mention spanning!

Why?

[Doug Neal]

I think Linux and Windows can both do this quite easily in software... but why bother? JBOD is the worst of both worlds when it comes to storage arrays. You have all the risk of losing everything if one drive dies, without gaining the performance benefits that RAID 0's striping gives you. Hard disks are cheap enough for a 2TB RAID 10 array to be affordable.

Yes this was quite a predictable comment, but someone had to say it..

Re:Why?

[Dan Stephans II]

Use a JBOD enclosure to present the physical devices to the OS where you can do your own logical volume creation and management (such as with ZFS). Just because he wants a JBOD enclosure doesn't mean he's going to use the physical volumes individually.

Re:Why?

[caseih]

No that's not correct. JBOD is just that. Just a bunch of disks. Has nothing to do with redundancy (or lack of redundancy). What you do with them is completely up to you. You can implement a RAID-Z with them on solaris (which is actually faster on my Enterprise-class disk array than the built-in RAID-6 in hardware!), Linux RAID-5, RAID-10, or whatever. Except for issues of battery-backed caching, I have come to the opinion that for most low- to middle-end storage needs, a large JBOD and software RAID is the way to go.

Re:Why?

[Vancorps]

I see that as a good method for tier 2 storage and definitely for backup tier storage.

I wouldn't want primary storage to use that method for obvious reasons but I haven't really played enough with alternatives to Raid 6 so you've given me something to try! Thanks

Re:Why?

[Rockoon]

The main reason people choose JBOD is because they have a bunch of differently sized drives, which are not well suited for redundancy or striping.

In my surviving collection of misc drives, I've got a 40 gig (8 years old), a 200 gig (5 years old), and a 500 gig (9 months old)

There isnt any concievably usefull redundancy method using these, but I can treat the entire lot as a 740GB backup drive..

If its for a home media server, backups and redundancy probably isnt a serious issue.. and performance definately isnt.. capacity would be the only real issue..

Re:Why?

[Jason Pollock]

In my case, it's because I don't care if I lose the data. They're rips of DVDs/CDs that I own, so 1 DVD represents 5minutes of time. In a lot of the RAID setups, if you lose a disk, you lose the entire RAID. In others, if you lose the card/motherboard, you lose the entire RAID.

In that situation, the frustration represented by losing the entire array when a disk (or card) bites the dust is a lot higher than the performance benefits, or the supposed reliability benefit.

Remember, in a consumer environment it can take _weeks_ to replace a drive under warranty.

Heck, I use drive failures as a method of culling media on the server. :)

Re:Why?

[adolf]

large JBOD and software RAID/quote

Isn't that redundant? What does the above text specify which could not be concisely written with just the words "software RAID"?

Re:Why?

[caseih]

No, it's not redundant. JBOD has nothing to do with RAID. JBOD is just raw LUNs (disks) over a bus. You can put them together however you want. Software RAID is the most common thing to use with JBOD-exported LUNs.

I'm not surprised that you haven't made the distinction, though. JBOD is an enterprise term and tends to be used when working with large external (Fiber Channel) disk arrays, either as a mode of operation, or meaning a chassis of disks without a hardware RAID backplane, over a SCSI bus or Fibre Channel.

Most home users with internal disks don't ever technically deal with JBOD in the enterprise parlance. An eSATA adapter and a chassis of eSATA disks is very similar to JBOD, although it's accessing disks over individual eSATA busses, rather than a fiber bus. But I think the idea is the same.

With my mythtv system, I never bother with RAID or LVM anyway. I just add disks individually since I'm more worried about raw space than data loss. If I lose a disk I only lose whatever programs were on that disk. Eventually I suppose I should do a proper RAID-5, but I only have 2 disks right now and can't be bothered I suppose.

Re:Why?

[adolf]

Well, yeah.

But make no mistake: I know the terms. And it's got nothing to do with the bus used, or whether the disks are multiplexed with LUNs.

The words "software RAID" make it damn near implicit that there's Just a Bunch Of Disks attached. Therefore, I continue to suggest that software RAID+JBOD is redundant terminology, and that just saying "software RAID" is perfectly descriptive. (What else would you be software RAIDing[1], after all?)

To use a car analogy: It's like saying "I have a car, with tires on it." It's implicit that the car must have tires, or it wouldn't be a usable car. So, the explicit specification that the car has tires is redundant.

If you still don't understand, then please look up the definition of "redundant" with regards to common English parlance.

[1]: The only exceptions I can think where one might use software RAID without JBOD are loopback-mounted files (whether over the network, or local on one disk), or perhaps something else which is beyond "Just a Bunch Of Disks," like a software RAID array of iSCSI disks located within one or more additional, otherwise-independant, machines.

Re:Why?

[caseih]

Ahh bring out the car analogies! But if I buy tires does that mean I have to put them on a car? I can see your point, though. Software RAID does imply JBOD in some form or another.

I'm not quite sure that most people would think of "loopback-mounted files" as a use for JBOD, but I can think of quite a few uses for JBOD without RAID. The MythTV thing I mentioned, for one. In some cases you might want to throw a disk in the array and export it as a LUN for use as a virtual machine hard drive via a fiber channel switch. Or maybe you have a backup system where each disk should be able to stand alone for recovery purposes. Or maybe you need some temporary space so you throw a 1 TB drive in the array and mount the LUN on some server temporarily (have done this on several occasions).

Re:Why?

[atamido]

No that's not correct. JBOD is just that. Just a bunch of disks. Has nothing to do with redundancy (or lack of redundancy).

This is incorrect. JBOD is similar to RAID 0 without striping, allowing one to use disks of dissimilar size. There are some RAID controllers that will incorrectly refer to presenting physical drives directly. However most RAID will correctly present a JBOD as a single logical volume.

Please refer to the Wikipedia article on RAID.

Re:Why?

[adolf]

Right. I agree with each of your points: JBOD doesn't imply any sort of RAID. However, software raid does imply JBOD.

That you have tires does not imply that they're mounted on a car. However, if you have a car, it's implied that it has tires.

I'm not quite sure that most people would think of "loopback-monted files" as JBOD, either (unless they were working on the md bits in the kernel or something). Which is why I used it as a counterexample. It's such a corner case that nobody sane would call it JBOD (even if, in Unix, everything is just a file anyway).

At the end of the day, there are no practical or common examples of "Software RAID" existing without JBOD that I can think of, which is why I still continue to say that "software RAID" implies JBOD so strongly that it is redundant.

It's a minor point. Hardly even worth discussing. But I still believe it to be true. I mean, if you walked around saying "Yeah, I just got a new car, AND it has tires," everyone within earshot would look at you funny -- of course the car has tires. Like saying "Yeah, I just got my new 5TB software RAID array up running AND it's JBOD" -- of course your RAID array spans a bunch of disks. What else would it span? Punch cards? Paper tape? CD-R? Come on.

Re:The best ESATA isn't really ESATA at all.

[drsmithy]

You're better off with an SAS external enclosure and a SAS card with external connections. These can be expensive, but will pay for themselves quickly with the lack of extra management.

What management ? You get an eSATA chassis with a port multiplier, slot in some drives, and run a single cable to the eSATA port on the computer. "Management" doesn't even come into it.

It's a home media server. In what was is SAS even remotely justified ?

If you need more than ten disks, go for cheap SAS

[Robotbeat]

You can get an external (4-port, but acts like one big 1.2 GiByte/s pipe) SAS RAID card for less than $500 that will allow you to make multiple RAID sets of up to 32-disks in a set using true hardware RAID 5,6,10, etc. You can even get a battery backup unit for the RAID card cache for $100 (priceless on critical DB systems).

An external SAS card allows you to connect over a hundred drives through one connection using SAS expanders (some cards support up to 256 devices). Some external SAS RAID/JBOD cards have two SFF-8088 connections, for eight SAS lanes total. That's 2.4 Gigabytes/sec raw. At that rate, it's your PCI-e bus that's usually the bottleneck.

A lot of SAS expanders are expensive, but Chenbro has some ones for $300 that spread one x4 lane SAS cable into 24 or 32 cables, plus they can be daisy-chained for more storage. Then, buy a nice 24-slot Supermicro 4U chassis with dual-redundant power. That's a little less than $1000. All you need is the Chenbro expander in the chassis, no need for a motherboard.

If you're really cheap, you can use a cheaper $150 external SAS JBOD-only card, but hardware raid really is a must if you have a lot of storage. Plus, a hardware raid can use write-back cache, since it has effectively non-volatile RAM using the battery backup unit. And no, a UPS is NOT a replacement for NVRAM... Has your system ever crashed for any reason or hung for any reason? I've never had a RAID card hang or crash.

So, basically, besides the external SAS card, you have:

24-slot chassis with redundant power: $1000
chenbro SAS expander: $300
cables: depends

That's about $60/slot, plus you have redundant power (and an upgrade route to dual-redundant controllers). You can scale this to hundreds of terabytes, too. Over a petabyte if you have multiple controllers (with raid array rebuilding on one card not affecting rebuilding on another).

Re:If you need more than ten disks, go for cheap S

[Robotbeat]

BTW, you can use SATA disks with this SAS setup. Also, this is hot-swap.

Re:Just nitpicking, but...

[TheRealMindChild]

Except when your backup server uses RAID...

Re:Just nitpicking, but...

[PitaBred]

The RAID itself is not backup, though. The backup server is the backup. How it stores the data is immaterial.

The Rosewill RSV-S8

[UserChrisCanter4]

The Rosewill RSV-S8 is pretty much exactly what you've described. It's an eSATA enclosure with 8 drive caddies, a power supply, and a fan. It presents the drives to the system as JBOD or one of the various common versions of RAID (implemented in software, I assume). Ignore the comically inflated MSRP; it's $300 on Newegg. It ships with its own eSATA card for compatibility purposes, but I assume it would work with any eSATA adapter that followed the proper specifications. There's also a five drive version available for about $100 less, give or take. I can't speak to the reliability or ease of use, but this sounds like it will fit your requirements.

Re:The Rosewill RSV-S8

[Zerth]

This is a 5 disk eSATA for $180. Appears to be similar (Silicon Image bits, single cable w/port multiplier, etc)

Re:The Rosewill RSV-S8

[lanner]

Unless I am mistaken, this device is a re-branded Venus T5 storage enclosure, made by AMS. With Linux, it works well. I've read the FreeBSD doesn't currently support SATA hub/switches (or whatever it's called), though that may be old info. The AMS device comes with Windows software, but I've been told it sucks.

Re:The best ESATA isn't really ESATA at all.

[conspirator57]

and you do this until you get to ((3.0 Gbps) /(8 bits/byte)) / (best case disk read speed in MiB/s) number of disks. Then you floor that to preferably a power of two or at least a natural multiple of a power of two.

This maximizes your bandwidth to the resulting filesystem.

e.g.

3Gbps = 402653184 bytes/s ~= 400MiB/s

i get some seagates with 75 MiB/s max sustained read.

400/75=5.3 -> 5 disks

5 is not divisible by a power of 2 (duh, its prime)

4 is the closest.

sector size is listed at 512bytes, so ideal stripe size is a multiple of 2KiB. This keeps your IOs clean and the disk cache happy and unfragmented.

enjoy a relatively optimized JBOD.

Re:The best ESATA isn't really ESATA at all.

[LotsOfPhil]

5 is not divisible by a power of 2 (duh, its prime)

Interesting way to see if 5 is divisible by 2. I think the same way:

What is 9 times 7? Well, it has to be less than 70 because 9It has to be odd since 9 and 7 are odd.
61 and 67 are prime so they're out.
65 and 69 are square-free and 9 is a perfect square.
So, 7*9 = 63.

(apologies to the actual (and funnier) version that I read but can't quite remember in Symmetry of the Primes)

Re:The best ESATA isn't really ESATA at all.

[LotsOfPhil]

Despite being at the forefront in almost all areas of number theory, Kummer was renowned for being very poor at elementary arithmetic. (A number theorist who was poor at arithmetic!) One story has him standing at the blackboard during a lecture, unable to compute 7 times 9. One mischievous student suggested 61, so Kummer wrote this on the board and started to continue. Another mischievous student shouted out that it was 69 not 61. At this, an exasperated Kummer, said "come on gentlemen, it can not be both". Later, it was rumoured that he told colleagues, he should have known the answer since it couldn't be 61 or 67, because 61 and 67 are primes and it couldn't be 65 because 65 is a multiple of 5, and he should have realized 69 was too large because 7 times 10 was only 70, so the only odd number left in the sixties was 63.

Be careful about your hardware and software

[ballyhoo]

If you're going to to this, you really need to be very careful about your choice of hardware and software. You need to avoid anything which isn't AHCI 1.3 compliant, as previous versions of the AHCI specification defined only a single FIS register per port, which effectively means that the controller card has to serialise all commands to the port multiplier. So even if you've got a port multiplier with a pile of separate disks, your throughput is going to be trash because the host operating system can only talk to a single disk at any one time. AHCI 1.3 fixes this and allows the host operating system to talk to multiple drives simultaneously.

You also need to be careful in your choice of software driver and operating system. Most of the free unix clones have some form of support for port multipliers these days, but this support is not really optimised towards high performance from sensible hardware yet. NCQ (native command queueing) is really important for performance here. I'll guess that with Windows drivers, you just won't know in advance, because the drivers aren't open source and you just can't tell what's going on inside them.

As previous people mentioned, it's important to configure multiple disks like this in some form of redundant mode. If you have a single volume spread across 5 disks, your risk of failure is going to be 5 times more likely than for a single disk, and the consequences of losing that data is 5 times worse than that of a single disk.

Re:Be careful about your hardware and software

[m.dillon]

If you use an AHCI controller it has to have the FBSS (Fis-Based switching) capability, which is basically AHCI1.3 as long as the capability bit is also set. The AHCI driver must also support the feature. Very few AHCI controllers have this capability, it is very new. The AHCI driver must also support AHCI's NCQ (multiple tagged commands). Most AHCI controllers have NCQ support now but if the driver doesn't you'll be stuck with horrible performance. Your chipset and driver needs to support both NCQ and FBSS to get decent performance through a port-multiplier-based enclosure.

Other non-AHCI controllers can also do Fis-Based switching, most notably something like the ultra cheap Silicon Image 3132. However, you have to be extremely careful with these chips because they have some significant hardware bugs and if the driver software doesn't work around the bugs you will get a lot of data corruption.

Unfortunately, even with the correct combination of software and hardware and FBSS, enclosures which use SATA-based Port Multipliers are still very sensitive to error conditions. For all intents and purposes if a device has a problem, the entire port multiplier enclosure has to be idled to clear the condition. So while this stuff can be used for RAID-like setups don't count on performance being maintained if one the disks starts to go bad.

There are a ton of cheap PM (Port Multiplier-based) enclosures on the market which take SATA drives, a lot of them use Silicon Image port-multiplier chips (I forget the exact chip number). The chips work but are a bit quirky. Still, drivers can work around the issues and still get reliable Hot-Swap events. The main reason for using something like that is to be able to have a simple backplane in the PM enclosure so the drives can be hot-swap. Most of these enclosure chipsets do seem to support Fis-based switching. Everything in the chain has to support it or it can't be used.

Port-multiplier-based SATA hot-swap enclosures are the cheapest available. Prices will probably continue to drop since the hardware needed is minimal. It's basically two chips, a backplane, a power supply, and the hot-swap elements.

In terms of operating systems and drivers. Well, Linux seems to have fairly decent drivers. I just finished doing full-on drivers for DragonFly for AHCI and the Sili 3132. The Sili driver can do FBSS+NCQ. The AHCI driver can only do NCQ atm. Plus Hot-swap, of course. I think Solaris has a decent AHCI driver. Other OS's have drivers with varying levels of support (w/ regard to NCQ, FBSS, Port-Multiplier, Hot-swap, and PM-based hot-swap support), and varying levels of robustness.

Another way to go is with the more expensive SAS chipsets. These can generally handle SATA drives too and theoretically have a more robust hardware programming interface.

-Matt

Re:raid can help with backups

[ls671]

He NEEDS another computer on his network.

With only one computer/disk controller if one of them fails, all FS might end up toasted.

He also needs incremental backups, just overwriting a snapshot of you data is no good when you realize that you have just overwritten your data with corrupted data because your main computing is failing slowly.

Re:Just nitpicking, but...

[Vancorps]

I'll disagree strongly with you as a 100TB library costs about 70k versus a 60TB proper SAN which will run 100k. Both will provide me with what I need in terms of backup but they both have their drawbacks. The library will take me weeks to recover from while the SAN can keep me up and running with zero down-time. It matters if your company depends on being online.

Of course for the home, I'm a fan of backup to an online service as restore time doesn't really matter and then you don't have to maintain separate gear.

Re:If you need more than ten disks, go for cheap S

[pyite]

hardware raid really is a must if you have a lot of storage

No, hardware RAID is a bad idea. You're locked to a proprietary controller and a proprietary on-disk format. ZFS is a much better idea.

Re:The best ESATA isn't really ESATA at all.

[KonoWatakushi]

Make sure to find a port multiplier with FBSS (FIS-based switching) support. Also make sure that your SATA controller supports this feature. Otherwise, there can only be one outstanding command for all attached disks, and performance will be abysmal.

Re:If you need more than ten disks, go for cheap S

[zen_sky]

Jeez, why do I feel I like drank my slurpee too fast? Thanks for the info dump! :)

Re:If you need more than ten disks, go for cheap S

[Robotbeat]

ZFS won't give you good performance for a large array because your random read speed is basically limited to the equivalent of one drive per raid set. That is unacceptable if you need performance:

http://utcc.utoronto.ca/~cks/space/blog/solaris/ZFSRaidzReadPerformance
"...adding more disks to a ZFS raidz pool does not increase how many random reads you can do per second."

Old AT (pre-ATX) case

[metallurge]

The old AT cases had a power supply with a mechanical power switch, rather than a soft-switch like ATX power supplies. Old AT cases and power supplies should be just about free, just strip out the old motherboard and you have a decent, inexpensive solution. Like someone else said, just get long SATA cables, and run them directly to the drives. You can bundle them together with zip ties periodically down the length, or use wire loom if you want something a bit neater. You may need molex-to-SATA power adapters, but those are very cheap and reliable. If you pick the right case, it will have plenty of drive bays and cooling capacity.

Or, you can use one of those 4_3.5"_drives-in-3_5.25"_bays solutions if you need even more space and cooling capacity beyond what is already in your case. Even a small mid-tower case should support at least 6 drives using one of these.

Pick up a spare AT power supply while you are at it, and you will have a very reliable, well-cooled, very cheap solution.

Re:Old AT (pre-ATX) case

[metallurge]

If you want to make it really neat, cable-management-wise, get some of those SATA-to-eSATA brackets to go between the drives and outside of the AT case.

Re:Old AT (pre-ATX) case

[cadu]

Cool solution, BTW, if you don't have an AT case and power supply, you can 'emulate' that hardware switch behavior even with an ATX case:

  1- Get the main 20/24-pin atx molex and use a clip or a small piece of wire to patch permanently the only _green_ (atx power sens) wire to any black wire (GND) or if you don't care about reusing the power supply, ditch the molex, insulate everything else and twist the green and one black wires together, and insulate.

  2- If you have any cruft inside this case (old cdrom drives/motherboard) remove everything and put only the hard drives for extra air flow :)

  3- wire with sata-2-esata adapters or very long sata cables.

  4- Turn on your 'JBOD' with the power supply's back power switch (AT-like behavior)

  5- ????

  6- Profit??

Re:Old AT (pre-ATX) case

[cerberusss]

Old AT cases and power supplies should be just about free, [...] you will have a very reliable, well-cooled, very cheap solution.

Two years of constant running could mean that a standard enclosure, consuming less power, is actually cheaper.

Re:Old AT (pre-ATX) case

[metallurge]

First, AT power supplies are switchers, so they are reasonably efficient.

Second, the AT power supply is going to work out to be sized appropriately for a case full of drives, so it will be operating in its sweet spot efficiency-wise.

So third, I am not sure what basis there would be to assume that standard enclosures are going to be any more energy efficient. A case full of drives is going to consume significant wattage regardless, and an AT power supply is a reasonably efficient way to provide that wattage.

Re:If you need more than ten disks, go for cheap S

[pyite]

ZFS won't give you good performance for a large array because your random read speed is basically limited to the equivalent of one drive per raid set. That is unacceptable if you need performance

Cheap, reliable, fast: Choose two.

Cheap + Reliable: RAID-Z and cheap drives
Cheap + Fast: Stripe on a non-ZFS filesystem
Reliable + Fast: ZFS mirrors on good drives. Go a step further, add L2ARC on SSD (readzillas).

You can't have your cake and eat it too. That said, RAID-Z and RAID-Z2 perform quite well for most people. I like the data integrity ZFS offers, but that's just me. I'm done with traditional filesystems and volume management.

Re:The best ESATA isn't really ESATA at all.

[ppbimix]

You're better off with an SAS external enclosure and a SAS card with external connections. These can be expensive, but will pay for themselves quickly with the lack of extra management.

What management ? You get an eSATA chassis with a port multiplier, slot in some drives, and run a single cable to the eSATA port on the computer. "Management" doesn't even come into it.

It's a home media server. In what was is SAS even remotely justified ?

Forever to go ! www.hjtj.ru

Re:The best ESATA isn't really ESATA at all.

[Anonymous Coward]

I don't understand why people simply don't evaluate by doing something like this: 9999 x 4 = 10000 x 4 - (1 x 4) doing this in your head is much simpler than moving digits all around.

Re:Just nitpicking, but...

[growse]

Why RAID your backups? Why bother?

pc-pitstop.com

[kenh]

At $WORK we just got a nice 8 bay rackmount eSATA chassis from them - dual/redundant power supply two quad-port SAS connectors, about $895, $679 for single power supply version. We bought it with 8x 1TB SATA HDs and an Areca RADI card with cables for just over $2200. (it is available as a chassis without cables, cards, or drives).

Well, this is fun

[Master of Transhuman]

OP asks questions about external eSATA enclosures, the entire first page of responses is an argument over whether RAID is backup... /....

Here's an ON-TOPIC RESPONSE! Horrors! Take away my EXCELLENT KARMA for this breach of /. protocol!

I have a client who needed backup for a lot of big video files. We bought an enclosure from PC Pitstop, eight bays each holding 750GB SATA hard drives (1TB wasn't really around last year when we got it) attached to two eSATA cards in the PC controlling the enclosure. We spent a month futzing around trying to get the enclosures to be seen. I forget who made the eSATA controller cards but they sucked - or the enclosure chips sucked.

So we turned to Burley, the guys who make enclosures for Macs mostly, but they work with PCs, too. These guys know their stuff. They told us not to use OEM hard drives in enclosures because some OEM drives you buy are dumped on the market and don't QUITE work with enclosures. They said use retail hard drives only. They also sell very good controller cards. The enclosure we got from them has worked fine for the last year and a half until last week when one of the drives went dead - no surprise. They aren't cheap, but they are well made and support is very good. I had both email and phone conversations with the Burley folks and they provide good support.

We also in the last couple months bought two MicroNet 4-drive eSATA enclosures with 1TB drives from Newegg for use on a Mac Pro. That was a huge mistake, since the drivers simply weren't seen by the Mac at all. Apparently MicroNet didn't bother to test the drivers when Mac OS X 10.5 came out and couldn't be bothered to provide support for that. So we attached the enclosures to a Windows PC and they work OK, although occasionally one or more of the drives will disappear and generate "drive not ready for access" messages in the Windows event logs.

Later, we decided to use those enclosures for iSCSI storage served up to the video lab. So I took one of the video lab PCs that were being replaced by iMacs and installed OpenFiler, the open source storage server run on Linux. The latest Rpath Linux kernel saw the drives and the enclosure no problem. I configured the iSCSI setup and everything seems to be working fine. And interestingly, none of the drives have gone offline like they did with Windows - which means it was Windows fault, not the drives. So now I can install an iSCSI client on the two iMacs - except Apple doesn't HAVE a Mac OS X iSCSI client, once again demonstrating how Apple isn't ready for the enterprise, since Linux has had them for years - fortunately there's a free Mac iSCSI client from another company - and serve up 1.8TB of iSCSI storage to each iMac.

So my advice is: choose your enclosures and the drives in them and the controller cards carefully. Take notice of what Silicon Image chipsets are involved, since SI pretty much dominates the market for those things and they're not the smartest tech company in the world. Make sure you get retail disks for use in the enclosures. Make sure you can return what you bought for refund or replacement because this stuff is not yet "set and forget".

Re:home array backups?

[metallurge]

Looks to me like there are two convenient, relatively affordable options. First, if you have access to old hardware, you can rebuild a spare computer into a cheap backup server with a suitable number of disks (linux, samba, software RAID or JBOD if you like living dangerously). Keep in mind, it takes a looong time to move 5 Tb of data over the network.

The second option is directly-connected individual drives, either permanently mounted in external cases (recommend eSATA) or just bare drives which you can dock one at a time using something like the Vantec Nexstar SATA Toaster Dock (NST-D100SU).

There are advantages & disadvantages to both methods. Both are cheaper per Gb than any other reasonable solution (tape, optical, etc).

Depending on how much and how frequently your data changes, using a hybrid system with a full backup set to HDD and incrementals to DVD-ROM may also be practical.

Re:The best ESATA isn't really ESATA at all.

[CrashNBrn]

Is this really as easy as you claim. I will admit to not being familialr with this side of computer hardware. NAS/SAS/portMultipliers/et al... even after reading http://www.addonics.com/products/host_controller/tutorial_pm.asp , I still don't see how it can be management free.

I have 4 drives in my main box atm:
8G Seagate c.1999
160G Seagate SATA c.2005
300G Seagate SATA c.2006
500G WD SATA-2 c.2007

These are not raided, primarily DATA drives. At times data will be shuffled around from Drive to Drive or reorganized.
The claim here on slashdot is that RAID is not a backup... So then, what is? If I hook up my unused 500G drive - and start using that for backup, it will get full. So I buy another. Now I have to know where the backups go. And if I move data around in my main chassis then the previous backups will be inconsistent. IE stuff that was on Backup drive 1 which used to match what was on Main drive 1... would no longer be the case as a lot of data may have been moved from MainDrive 1 to Main drive 2. So an unmanaged backup would wind up with the same information on Backup#1 as Backup#2 - or you would wind up having to move data around between the backup drives as well - to match what you have done on the main system.

I've read through most of this thread, and I still don't really see any actual answers to the OP's question. Where is the software that can remove the headache of managing JBOD's that aren't raided? Does the hardware somehow take care of this - it really doesn't seem to from the information I have been able to acquire.

Re:The best ESATA isn't really ESATA at all.

[Techman83]

SAS can be justified, You can buy a cheap 8 port card, that can take SATA disks and it will be pretty darn quick too.

Re:The best ESATA isn't really ESATA at all.

[captrb]

Maybe it isn't justified for this purpose, but SAS doesn't have to be exponentially more expensive. First, why bother? Well, eSATA sucks compared to SAS. SAS is faster, better error reporting, more options, better drives, better cables, better controllers, and far more expandable in a huge way. Now, price... Seagate has 7200rpm ES2 SAS drives that are only a few bucks more than SATA drives. If neccessary, you can add a couple of 15k drives for intensive tasks (video editing?) alongside 7.2k drives for archival. All that without slowing the bus to SATA speeds (you can mix SAS & SATA, but it slows the bus to SATA speed). It can be a huge bargain. To the grandparent post: just buy the 4U supermicro chassis (plus JBOD power control board) that has a backplane with an expander (or two). These allow you to stack up and span out as far as your controller/bus can run. You could also just throw a motherboard in it and run OpenFilter or Nexenta to have a real storage server that does FTP, NFS, CIFS & iSCSI. Plus snapshots and such.

I found dedicated enclosures to suck

[Kjella]

Eventually I just bought myself an Antec 1200 and a MIST PSU with modular cables. Loaded it up with a SATA rich mobo + a small SATA card for 12x SATA data/power. Because of an earlier RAID accident because of poor warning setup (two disks failed with some time between, but I didn't notice the first one) I do JBOD and manual copies, but you could just as easily do software RAID - the "hardware" RAID on these aren't worth it anyway. That way I have a full Linux server I can use for whatever too.

Honestly, if I wanted more backup I'd probably get another. It's been rock stable, drives are all below 30C w/dust filters and if something fails I don't need to get the exact same RAID controller. Price was about same as most 4-bay SATA enclosures and if you pick the lowest end processor it doesn't consume much compared to the 12 drives anyway. It's thumbscrews and not hotswap but I can afford to take my server down while I'm fiddling with changing disks - like I said all disks seem to live happy lives in there though. Right now I got 3*1TB+2*750GB+5*500GB+250GB+160GB = 7410GB online. With modern 2TB drives I could do 24TB. Basicly, works4me.

Re:Drobo

[Anonymous Coward]

See the Drobos at the linked page. No eSATA, but perhaps you can get an eSATA -> firewire 800 or iSCSI (sp?) dongle. The best part about them is ease-of-use.

and the worst part is that their users all seem to be morons who don't realize they've been ripped off by slick marketing.

Probability

[dna_(c)(tm)(r)]

A single disk is more risky than I would like. Especially since it's offline, it can fail without warning [...]

Your offline disk is used much less than the disks it backs up. Much less likely to fail. But a single backup disk is not a good idea for that reason. Rotate a few of them.

Ever wondered why airliners seem to move from 3 or 4 to 2 engines? Because the failure rate is more or less independent of the size of those engines (similar number of components). I know, it's counter intuitive.And cost and maintenance is involved too...

Now suppose a single disk has 70% probability of failing between 4 and 6 years or 5 +- 1 year (source of statistic: Library of Thin Air), then

• A or B or both fail: P(A or B) = P(A) + P(B) - P(A and B) = 0.70 + 0.70 - (0.70 * 0.70) = 0.91
• A and B fail: P(A and B) = P(A) * P(B) = 0.49
• with 3 disks, A or B or C fail: P(A or B or C) = 0.976

So the probability of you needing to do something is higher when two disks are involved, if you don't check regularly in that 2 year period, you have 49% chance of BOTH disks failing.

There is another problem. These calculations are for the case where the failure of both disks is independent. Since they are in the same RAID setup, spike current, wear and tear, overheating etc. will follow the same patterns and catastrophical events are the same so chances for them failing within a short time apart are grossly underestimated. The airliner metaphor: the A-320 that crashed in the Hudson: one flock of geese, two engines...

Now, what happens if we look at a very small probability band (10%, same variance or standard deviation as in the 70% case ): 5 years +- 29 days (365*0.0797)

• P(A) = 0.10
• P(A or B) = 0.19
• P(A and B) = 0.01

The chance of both failing in the same 2 months is not huge, but still 1%, that's not entirely negligible.

My advice? Use lots of disks, offline, swap regularly, test from time to time and if you still have a budget left (corporate budget) replace each single disk with some RAID setup if you must...

Re:Probability

[TheLink]

> Ever wondered why airliners seem to move from 3 or 4 to 2 engines? Because the failure rate is more or less independent of the size of those engines (similar number of components). I know, it's counter intuitive.And cost and maintenance is involved too...

Previously they needed 4 engines because the engines weren't as powerful and they weren't as reliable. Compare how much thrust the original 747 engines provided with the latest engines.

The thing is modern two engine jets are cheaper to maintain and seem to use less fuel than the 4 engine ones (not sure why they're more fuel efficient, maybe the 4 engine jets are older designs?). As for safety, they now can have planes with two engines where if one fails the plane can still fly safely for about 180 minutes and land, and the individual jet engines are more reliable nowadays. Not sure if 2 engine jets could do that in the 1960s :).

But everything being the same, a four engine design is safer than a two engine design, just more expensive. You can climb with 75% power. Climbing is harder with 50% power :).

Re:Probability

[dna_(c)(tm)(r)]

But everything being the same, a four engine design is safer than a two engine design, just more expensive.

My point was that it isn't. You have a higher probability of having one or more engine failures on a four engine design than on a 2 engine design. It's not a matter of opinion, that's a mathematical fact.

BTW a four engined aircraft losing 2 engines on the same side generates a huge amount of torque around the vertical axis, which might be more problematic than loosing 1 engine on a 2 engine design.

The other point was that in general close proximity subjects the engines to the same external factors (as with disks) so that even 4 engines can shut down simultaneously as in this case.

Re:Probability

[sjames]

I don't mind the higher probability that I will need to do something, disks are cheap and it doesn't take long to swap one. What I mind is that without RAID, when I do need to replace one, whatever was on it is gone. I use RAID6 so that I can lose any 2 disks and still recover my data without any heroic efforts. The primary is RAID 6 as well, so I would have to have a disaster on the primary AND lose 3 disks in my backup within the same window to end up SOL. It's not impossible, but then there is no scenario at all where data loss is actually impossible.

When feasible, I use disks of different ages in my RAID. That avoids having too much of the array at the end of the bathtub curve at the same time even when I don't know the exact shape of that curve. Necessarily, that means the disks are not from the same batch.

I'm not worried about the probability of losing multiple disks in 2 months, I'm worried about losing 3 disks within a day or two.

I would have to ignore a LOT of error reports to go two years without noticing.

Re:Probability

[sjames]

The failure probability for one engine matters because engine repair is expensive. If it were as cheap as swapping out a bad SATA drive, nobody would care.

What people care about far more is having more failures at the same time than redundancy allows for. That's when things go really bad.

Since we were talking about disks and not aircraft engines, torque is not a factor. Analogies are great, but only if you don't let them lead you astray. As long as I have N+2, losing any 2 drives is not a disaster (but certainly is an urgency).

Yes, there exist situations where multiple redundancy can fail as well. However, I refuse to spend trillions building a DR datacenter on Mars just in case a CME burns the Earth.

When I am maintaining a backup system, the primary and secondary being geographically seperate takes care of black swans like massive power failures (BTW, I KNOW that my hardware can stand 208V suddenly appearing on the 110V outlet because it actually happened once).

Archival storage is another problem set entirely, but that can avoid black swans by not being powered on most of the time (though periodic power up and testing is highly advisable).

Re:Probability

[dna_(c)(tm)(r)]

RAID 5 loosing any 2 disks is non recoverable, RAID 6 loosing any 3 disks is non recoverable, isn't it? If you have 4 disks and group them in 1 RAID 5 that's worse than two RAID 5s of two disks.

Re:Probability

[sjames]

That would be why I use RAID 6. It is possible to keep the window for failure to a minimum by including a hot spare as well.

With two disks, you can't do a proper RAID 5, you need RAID 1. The better choice though is probably use all 4 disks in a RAID 6.

Re:Probability

[TheLink]

Sure, but having one or two engines fail on a four engined plane is not as serious as having one or two engines fail on a two engined plane.

The FAA still believes 4 engined aircraft are safer than twin engined aircraft, hence their extra requirements for twin engined aircraft to fly long distance routes (ETOPS). Maybe you can go show them why they are wrong.

Your example does not disprove my claim that a four jet is safer than a twin jet. Since a twin jet is likely to also have all engines shut down in your stated scenario.

I'm not sure if a twin jet would have done better, in fact a twin jet might have not made it. In your example, after the 747 restarted a single engine out of 4, it helped to slow the rate of descent, giving time to finally start the other and thus climb and eventually make a safe landing (great flying by the crew).

The way I see it, let's just say the chance of an engine failing = 0.01

And 50% engines = OK , and < 50% = not ok. As per your link, the 747 could climb with two engines (even engines that had volcanic ash blown through them).

With 4 engines, the odds of any one engine failing = 0.04, the odds of any two engines failing = 0.0004, the odds of any three engines failing = 0.000004, the odds of all four failing = 0.00000001
With 4 engines, the odds of < 50% = 0.000004 + 0.00000001 = 0.00000401

With 2 engines, the odds of any one engine failing = 0.02 (yep half), the odds of both engines failing = 0.0001.
With 2 engines, the odds of < 50% = 0.0001 which is more than 0.00000401

Help fix my math if it's wrong.

If the design is 75% engines = not ok, then 4 engines is worse. But they don't design planes like that.

A 747 can climb and cruise with 1 dead engine (and lose maybe about 10% range). Lots of 747s lose an engine and keep flying. A BA 747 plane flew from LAX to MAN with just 3 engines (the FAA and press made a lot of noise about that one).

Re:Probability

[dna_(c)(tm)(r)]

RAID 5s of two disks

My stupid mistake.

Re:The best ESATA isn't really ESATA at all.

[drsmithy]

Is this really as easy as you claim.

Yes.

I still don't see how it can be management free.

All a JBOD shelf does - by definition - is pass through the drive connections to the host. Ie: plug 3 drives into your JBOD, your system sees 3 drives. Plug in 5 drives, the system sees 5 drives. There's no management because there's nothing to manage.

Where is the software that can remove the headache of managing JBOD's that aren't raided? Does the hardware somehow take care of this - it really doesn't seem to from the information I have been able to acquire.

There's no headache because there's nothing to manage. Your 5-drive JBOD shelf exposes 5 drives to the system. What you then do with those 5 drives (connect to a hardware RAID controller, use software RAID, treat them as individual drives) is up to you.

It appears you (and a lot of people on this thread) don't understand what a JBOD enclosure is.

Re:The best ESATA isn't really ESATA at all.

[drsmithy]

SAS can be justified, You can buy a cheap 8 port card, that can take SATA disks and it will be pretty darn quick too.

But you can buy a much cheaper eSATA card that will be more than quick enough.

Re:The best ESATA isn't really ESATA at all.

[drsmithy]

It can be a huge bargain.

Not in any remotely realistic home use (and even a fair chunk of professional use) scenario it can't. You'd be looking at a minimum 50% higher cost, for questionable (if any) advantages.

Re:If you need more than ten disks, go for cheap S

[TheRaven64]

Note that this only applies to RAID-Z and RAID-Z2 pools. ZFS also supports mirrored pools and stripped pools (equivalent to RAID-1/0). If you care more about performance than data integrity (e.g. for /tmp, possibly for /var) then a stripped pool might be a better storage model than RAID-Z.

Re:The best ESATA isn't really ESATA at all.

[beerbear]

Also nice: 1001 x 999 = (1000 + 1) x (1000 - 1) = 1000^2 - 1^2 = 1000000 - 1
it's even nicer for stuff like 51 x 49

WHS

[po134]

WHS, so you can only "mirror" data that matters.

Sans Digital FTW

[pak9rabid]

I picked up one of these guys for my backup purposes. I filled it with 5 1TB drives and set it up in a Linux software RAID5 config. It backs up all of my media that resides on an LVM volume. It's been working out quite nicely so far :). The port multiplier feature is very nice. I only have to run a single eSATA cable for the 5 disks.

CFI-B8283ER

[kobold2]

8-bay esata, raid 0,1,10,5, s.m.a.r.t., hot-swap, sparing, internal 300w psu, affordable.

Re:Just nitpicking, but...

[PitaBred]

The point I was making is that having a RAID is not backup. The fact that it's not the same machine is what makes it a backup, rather than a failure-tolerance scheme.

Norco eSATA crates

[Ktistec Machine]

For what it's worth, we've had good luck with the Norco disk crates like this one:

http://www.newegg.com/Product/Product.aspx?Item=N82E16816133023

These have 15 hot-swap SATA slots. Each set of 5 is multiplexed to one eSATA connector on the back of the crate. The crate comes with a PCI-X 4-port eSATA controller. We use the crate as Just a Bunch of Disks, but it can be also configured as a RAID array. At the price (about $800), it's very cheap per slot. We currently have two of these full of terabyte disks, and an older DS-1200 (12-slot) with a mixture of disks. They've been very reliable so far.

Re:The best ESATA isn't really ESATA at all.

[CrashNBrn]

> It appears you (and a lot of people on this thread) don't understand what a JBOD enclosure is.
Apparently ;-) I'm a coder, my hardware knowledge was what I needed to build my box and/or troubleshoot family&friends PCs.

Everytime I bought a "big" drive for backup, it wound up inside my case for Data duty instead. This thread reminded me that I really ought to stop putting that off.
Making more sense now, and looks like a number of off-the-shelf-solutions are all raid and extremely overpriced.
Thanks.

How does battery back up help?

[pestie]

I've often wondered this and have yet to see an answer to it. How does having battery back up on your RAID card's cache help anything when your operating system is probably doing a buttload of caching in system RAM? A crash or power outage is still going to throw that away, leading to a just-as-corrupted filesystem.

eSATA or Multilane SAS

[petree]

First, let me suggest you consider using a card with a multilane SAS (4x) connector, also called infiniband, instead of an eSATA connection. These connectors are just 4x SATA/SAS bundled into one, so each drive gets full bandwidth instead of pushing it all across one 1.5/3gbps eSATA connection. You can even by a bracket that will combine 4x internal sata connectors to make one multilane sas connector, these run ~$20, so you can use onboard sata for 4 of your external drives if you've got extra sata ports on your mobo.
That said, here's the solution I have running with ZFS RaidZ under OpenSolaris (also used under Linux + FUSE ZFS):
DatOptic EBOX-M - 8Bay Sata Enclosure (other EBOX/QBOX enclosure available with eSATA/MiniSAS/USB/Firewire instead of Multilane SAS)
4x WD 1TB RE2 'Green Drives'
Addonics ADSA3GPX8-ML (SilImage 3124 Based, Certified for Solaris, works in Linux/Windows...MAC?)

I kind of wish I hadn't spent the extra loot on a the 8bay enclosure, since I still haven't dumped a second set of 4 drives, but imagine someday I'll fill it out. The enclosure is simple, no hot swap, just insert bare drives into the bays, no sleds/mounting hardware required.

Re:eSATA or Multilane SAS

[petree]

I forgot to mention why I chose the more expensive WD RE2 drives instead of normal retail drives.

The firmware on most drives is designed to make multiple attempts when it fails to read data off the disk on the first try. Sometimes taking 20-30seconds before telling the OS about the read failure. In a traditional RAID setup (or RAIDZ w/ ZFS) you would prefer to have the disk report the read error immediately and let the RAID card/OS handle recalculating the data using the redundant copies. RE2/RE3 drives play nicely like this, most retail drives just try a dozen times...and while that happens, Linux/Solaris/whatever may think the drive is dead since it's not responding and report it as a failed, degrading the array.

Also, the RE2 drives have a 5yr warranty and are supposedly designed for the vibration of multidrive setups.

the user friendly solution

[edrawr]

Flash drives, lots of flash drives. Drag and Drop. Come on...

unRAID

[Coppit]

If you're willing to get your hands a little dirty, check out unRAID. You put the OS (linux) and software on a flash drive and boot from that. You have a big parity drive and a bunch of data drives that are just ReiserFS. The user share feature can aggregate all your files into one big virtual filesystem. When you run out of space you just pop another drive in, or pop out a small drive and put a larger on in, then wait for the data to be rebuilt from parity. You don't have to worry about your RAID controller dying, or a two disk failure (you'll still have the other disks of data). I built my machine for maybe a couple hundred bucks, and just added my 7th disk. I have smaller disks in there, but don't bother to remove them.