Terabyte Storage Solutions? 574
DeMechman asks: "As many on Slashdot may know, storage is one thing which you can never have enough of. Given the current situation with CD/DVD rot (Personally I can attest to a 10% attrition rate) hard drives in a RAID configuration seem to be a better and more economical solution. If you own more than fifty CD/DVDs, it can be a daunting task to find a file. I am wondering if anyone has found a hardware solution that can inexpensively be set up to handle 10 or more 250GB HDDs in a RAID configuration. Primarily, has any case manufacturer tackled this niche market yet?"
What's "inexpensively"? (Score:5, Informative)
http://www.apple.com/xserve/raid/ [apple.com]
Academic prices for:
1.00TB - $5399
1.75TB - $6749
3.50TB - $9899
It's not RAID, but ... (Score:5, Informative)
RAIC - Redundant Array on Inexpensive Computers (Score:1, Informative)
Many have (Score:3, Informative)
See page here. [apple.com]
Intel SC5200 5U (Score:2, Informative)
Easy these days. (Score:5, Informative)
I have a TB here, and rather than raid, I decided to do a nightly "rsync" mirror to a "yesterday" partition.
The two advantages of the nightly rsync over RAID are
easy to do with rackmount cases. (Score:5, Informative)
Terabyte Storage (Score:5, Informative)
The 160GB drives used to come with a Maxtor [Promise] ATA-133 card. Two of those will support eight drives. Not the most optimal arrangement because of the bus having two drives on each channel, but it doesn't seem to affect performance too much since it is striping the data across all of the drives. I'm assuming it stripes in order, so you'd want to stagger the drives such that 1 & 2, 3 & 4 are not on the same controller.
Output of df -h:
The cost to assemble something like this?
~ $600.00
8 x $70 for the 160GB drives
2 x $20 ATA-133 controllers
The biggest issue is that there is no easy way to back up the array. You could use RAID 6 and have two drives worth of parity info, but it still leaves you vulnerable to a catastrophic hardware (or building) failure.
Anyone have any ideas on how to back up 1TB in a home environment? i.e., not $3000 tape drives & $200 tapes
Re:What's "inexpensively"? (Score:4, Informative)
Re:What's "inexpensively"? (Score:5, Informative)
Go buy a Lian Li case, 8 x 200gb maxtor harddrives and a 3ware raid controller.
Controller $500
Drives $150 each
Case $150
Total for 1.4TB = $1850
With 400gb drives maybe $3000 for 2.8TB
3ware Controllers + Drive Friendly Case (Score:2, Informative)
Re:What's "inexpensively"? (Score:5, Informative)
50 CDs * 700 MB = 35 GB
50 DVDs * 4.7 GB = 235 GB
It would take 250 DVDs (all FULL!) to get you to that terabyte. But you want to put ten 250GB drives together... so you want 4 drives (for the space) and six drives for redundancy.
Expect to put down $5,000+. Or buy a 250GB drive and just store them on there. Buy two, and use the second one as a backup of the first. Total cost? $400.
If you're a home user - don't go overboard. If you're a corporate user that's just trying to cut corners (and therefore cost) then don't shortchange yourself (or your company).
relatively cheap raid boxes... (Score:2, Informative)
Ever Try External Hard Disk Enclosures? (Score:3, Informative)
Re:2nd Question - Backups (Score:3, Informative)
Just built one... (Score:5, Informative)
I went with Serial ATA for a couple reasons:
1) It's cheaper and has more capacity than SCSI;
2) Cabling is not a mess as it is with regular IDE (if you've never seen serial ATA cables, the first thing you will notice is that they are small!);
3) It can hotswap, unlike regular IDE;
4) It's not that much more expensive than regular IDE.
I custom-built a 3U server from InterProMicro. [interpromicro.com] They are a small (local if you are in the Bay Area) SuperMicro reseller that does great work. (If you need something, call and ask for Andy. Tell him Erica from Simpli sent you!)
The machine I specced out was as follows:
* 3U case with 8 hot-swap SATA drive bays;
* 8-port 3Ware 8506-8 SATA RAID controller;
* 5x250GB SATA drives in a RAID-5 array;
* Dual Xeon processors.
The 5 drives give you 1TB of storage, and expanding up to 8 gives you 1.75TB. I would also recommend a separate mirrored SATA 10KRPM array for the OS if you want really fast speeds.
This whole solution (Xeons; 5 drives; 3U case) cost just over $3000... which is pretty reasonable for 1TB of network-accessible storage. Interpro has solutions that go up to 24 SATA drives [interpromicro.com], which at 250GB each gives you an ungodly amount of space (5.75TB, if my calculations are correct.)
My suggestion is to go with a niche server builder like InterproMicro over Dell or Compaq or any of those guys. You can get the same high quality from a custom manufacturer without paying the steep brand name price from a larger manufacturer. As for the drives, any time the goal is "as much space as possible", SATA should be your first choice.
Good luck!
Re:Many have (Score:4, Informative)
Lets see, 5 200 MB drives at $120 = $600 + another $600 for the case, MB proc etc... $1200 for a terrabyte server.
I haven't looked (you can do that) but I bet there are plenty of stand alone raid units of that size for maybe twice the DIY price and that is still HALF the price of Apple.
Now THIS is informative!
Good solutions still cost a reasonable amount (Score:5, Informative)
-What RAID level you want (5 usually requires better hardware)
-Whether you want hardware RAID (I strongly recommend this) or soft RAID
-How much redundancy you need (Battery backup cache? Redundant controllers? Hardware environmental controls?)
If you are looking for good pci cards, I would strongly suggest a card from 3ware [3ware.com], and a card from a place such a Seagate [seagate.com]. Getting a super-duper cheap card when terabytes of data are on the line is just fundamentally stupid. You can save some bucks now, but be ready with your next Ask Slashdot: "How do I recover data from my dead RAID?" Seagate now has a nice 5 year warranty, which match well with good quality and reasonably cheap drives. Look at some of the SATA drives like the Barracuda. However, any decent quality drive maker can work. If you have even more money, you can look at some of the things offered by places like StorCase [storcase.com]. A larger initial investment can become cheaper as you scale up the cheap harddrive count, and it can be a good thing in the long run. Obviously, the more time you are willing to invest doing things yourself, the cheaper you can get to some extent vs premade items. However, no support as well.
Do read up on some of the fundamentals of RAID: Everything you need to know (and lots you don't) is probably at least mentioned in the PC Guide [pcguide.com] on RAID. Look through that. Things like hot swap and hot spares are important to understand. Finally, you should remember to check compatability. Unfortunately, I for instance have not been able to find much of anything in the way of controller cards that is compatable with OS X (except the obvious, the XServe RAID). So I have something set up on a BSD box in my server closet that I then link to, more like a storage appliance. Happily, the 3ware cards and many others are now compatable with a wide variety of *nix and BSD flavors along Windows, but do check to make sure.
Last but not least, remember this!: RAID is *not* a backup solution, but an highly redundant onsite storage system. Have another form of backups, even if it is just a RAID 1 off site, or DVD-Rs, or something. If a disaster happens (thieves, fire, nuclear destruction, John Ashcroft) on site storage won't save you.
Off the shelf or build yourself? (Score:3, Informative)
If I were going to do it I'd build it my own by combining a nice case [rackmountpro.com] and a 12 port 3Ware controller [rackmountpro.com] with whatever server configuration and SATA drives I wanted to get.
Re:What's "inexpensively"? (Score:5, Informative)
12 x 3.5 bay" Tower [lian-li.com]
3 Ware 12 drive RAID card [3ware.com]
Drives can be found at PriceWatch [pricewatch.com]
I I havn't calculated the per MB cost of all the large sizes. someone with more time please do this.
What will make this perfect is removeble drive kits (They require an external 5.1/4" bay for each 3.1/2" drive. Some even have little activity LEDs) and a server case with 12 external 5.1/4" bays.
we made LOTS of 1.3 TB boxes at about $2000 each (Score:5, Informative)
Rip all your CDs as FLAC [sourceforge.net] so that (1) you never have to rip them again (it's lossless), but (2) it's half the size of saving WAV files
At least that's what we've done with our 68,000 [cdbaby.com] CDs we have here.
Re:Terabyte Storage (Score:5, Informative)
8 Firewire drive enclosure: (i have the 4 drives version).
$600. http://www.cooldrives.com/fi80013oc5fi.html
$136
$700 = Hardware for a Linux machine as correct file server
= $2930 for 2TB of raw space, 1.5TB Of raid 5 with an hot spare, or 1.75TB of raid five with no hot spare.
You got yourself a nice fileserver for home usage... install that with mythtv and you're up for hours of video....
Re:Easy these days. (Score:5, Informative)
> advantages of the nightly rsync over RAID are
Instead of only keeping a "yesterday" partition, use rsync to keep EVERY daily backup.
Rsync has lots of great options to make copies as hard links if they haven't changed and only copy changed files. That allows you to make daily full backups that only use the space of daily incrementals. Do that to a backup partition, then RAID-1 the whole drive over to a mirror.
That gets you full protection from hardware failure on a drive and user failure on your files.
Google for more details [google.com]
Re:What's "inexpensively"? (Score:5, Informative)
Some slashdot articles on some previous attempts:
Bulk Data Storage For The Common Man? [slashdot.org]
Home-brewing a 1.2TB IDE to Firewire Monster [slashdot.org]
Books on it:
Managing RAID on Linux [slashdot.org]
Even applicable controller hardware:
LSI Megariad 150-6 [lsilogic.com]
3Ware 9000 series [3ware.com]
And soon to be applicable storage hardware:
Hitachi Announces 400GB Hard Drive [slashdot.org]
Re:What's "inexpensively"? (Score:5, Informative)
Note about RAID (Score:2, Informative)
More info here, plus the ever-acidic jwz calling people dumbasses, dipshits, and more fun!
http://jwz.livejournal.com/368307.html [livejournal.com]
http://www.dnalounge.com/backstage/log/2004/07.ht
Re:Good solutions still cost a reasonable amount (Score:2, Informative)
Re:Terabyte Storage (Score:4, Informative)
According to Western Digital's site, a 250GB SATA drive pulls 12.8 watts when reading/writing and 9.5 watts on standby. I figure for 6 drives that's about 100 watts of a _good_ power supply's rating.
Here's our solution. (Score:2, Informative)
It uses a 9 external 5.25 bay case (enlight) with an Antec 550W power supply to handle the 12 drives (plus a seagate system drive in the internal 3.5" bay). This has worked very well.
We use Maxtor 300GB drives in one machine (RAID55) and have lost 5 of 20 drives we purchased in 6 months. The other uses Western Digital 200GB (RAID5), and we've lost 1 of 12 in a year. Manufacturer DOES matter. WD replaced our drive in days, Maxtor makes you jump through hoops and tries to deny the problem for a while, just to finally decide to replace the drive, then take 5-7 mroe days to get it to you.
All in all, these machines cost us under 7K each and perform very well. However, if I bought one today, I'd get 3ware's SATA card and Seagate's new 400GB SATA drives instead. Whoever said ATA cables are a pain was NOT wrong, and these drives would give much better performance.
My usual solution... (Score:3, Informative)
* SuperMicro motherboard (any of the newer ones, depend on your choice of architecture). Be sure to get one with PCI 133/64 and gigabit onboard.
* 3Ware RAID board(s).
* Chembro rackmount cases (they have a very nice one with 16 SATA hotplug slots with backplane and all)
* Don't go cheap on the power supply. You'll need at least 600W. I always go for redundant ones.
* 16 SATA disks of your choice (250, 120 or 80GB)
* Linux!!! (Be careful with fedora core2, it doesnt support nativelly the 3Ware cards - you'll need to compile your own)
Of course you could save about $1000 by using a cheap motherboard, chassis and PS. But it really pays off using the good brands on those.
By the way, you should always get an extra hard drive (or two). They will fail (sooner or later) and you don't want to be left hanging.
Re:What I did... (Score:5, Informative)
Re:What's "inexpensively"? (Score:3, Informative)
Re:Maybe LVM (Score:1, Informative)
Re:3ware Controllers + Drive Friendly Case (Score:5, Informative)
Our 48 Node beowulf has a
Anyway, that kind of load brought that head node (dual proc 1700+ MP) to its knees until we decided to rebuild it. Moving from the hardware controlled raid to linux's software raid completely resolved that problem.
Re:What's "inexpensively"? (Score:2, Informative)
Off the shelf, not home-made (Score:4, Informative)
Disclaimer: I work for a storage integrator, both are brands we sell.
Re:What's "inexpensively"? (Score:5, Informative)
Hopefully, they've changed this for the newer 7 series cards, but the 5 series are 'broken' this way.
Remember duty cycle and warranties (Score:1, Informative)
Also, remember that some drives nowdays don't allow for a 24x7 duty cycle. Given that the SMART diagnistics in the drives can tell quite a bit to the person examining your warranty return, don't try to 'cheap' your way through and then claim on warranty.
Re:What's "inexpensively"? (Score:5, Informative)
Oh, bullshit.
Linux software RAID1 is just as fast as several of the hardware RAID1 setups I've tested using Bonnie++ -- These are fucking fileservers, not renderfarms. The processor's sitting there doing jack shit anyway, and you're more than likely putting a P4 in there since you can't buy anything else with decent reliability. Throw in a decent GigE network card and your processor is STILL at 0% utilization. Make that a RAID5 with hot-standby drive and I would be very surprised if you noticed any difference in the apparent "feel" of the server compared to a hardware RAID solution.
Hardware RAID's okay but now you've got a proprietary format array with a SPOF (the RAID card(s)) -- sure you can keep spare RAID cards around but honestly unless you need every last bps on your network transfer and you've got your server so overloaded that SW RAID is impacting your performance you're just incurring extra expense. I am very happy that I can take any RAID array I have and throw it in another system should a motherboard or controller fail and I need the system up immediately. I'm very happy that LVM Just Works and works happily on top of software RAID. There's no issues and no extra question marks like there are with any hardware RAID "solution".
Want beeping? Write a script. Want email/phone/paging when something goes wrong? Write a script. Or use any of the monitoring and alerting systems you can find on Freshmeat (mon, nagios, etc.). Jesus H Christ, give your head a shake.
Oh wait, you're trying to build a performance system using an OS built for pushing pixels. Perhaps that is your biggest problem. Windows has its place, but high performance data transfer just isn't one of them. I guess if you've decided to spend a couple hundred on an OS license that gets you nothing you may as well blow another couple hundred to get hardware to go with it.
Somewhat off-topic, but... (Score:3, Informative)
Um... ever consider the mind-bogglingly simple solution of:
ls -R> ~/dvd.index/<disc_label> for each dvd
grep "<whatever_youre_looking_for>" ~/dvd.index/*
Re:What's "inexpensively"? (Score:5, Informative)
Re:What's "inexpensively"? (Score:2, Informative)
But Raid 5 storage efficiency follows (Number of Drives - 1) / Number of Drives) with an 8 drive RAID, that's 87.5% efficiency--and that's pretty dern good for a relatively decent fault-tolerant rig. That means that out of 1 Terabyte you lose the equivalent of 125GB, which isn't so bad for all of the benefits that RAID 5 brings, and it's a FAR cry from 40% usable as you claim. Hell, even RAID 1 (the most space-inefficient of all RAID configurations) is never more or less than 50% efficient.
Besides, you only need to snapshot the really important stuff (that can't be easily obtained from backup, and can't be easily recreated), it's not like you need to take 12 rotating snapshots of all your warez/porno/MP3 collection per day. This is all about a relatively cheap PERSONAL server.
You're doing it wrong (Score:3, Informative)
Two or three fairly normal PCs with STANDARD drive controllers, PSUs and HVAC.
Look, we're talking file servers here. 128MB RAM is gobs if you aren't running any other service on 'em. Pick and OS, any OS: 2000 gets you dfs, *nix gives you NFS. Both give you a homogenous networked file system.
So...
Standard case/PSU/cheapo CPU (Athlon mobile or Via or P3, for lower power consumption)/RAM - That's $250, maybe. Add another $20 for a gigabit NIC or two per machine.
4x 200GB drives @ $110 apiece (pricewatch shows $96 as the low price, but I'll go $110 for a little wiggle room)
So... something around $700 gets you
Buy three machines. $2100 gets you tons of storage and scads of redundancy no matter how you look at it.
This is the philosophy I use in setting up my file servers (now serving 6.5TB!). Over time I've added 3ware cards, upgraded PSUs and added gobs of RAM, but my basic starting point is a very modestly-appointed system.
Re:It's not RAID, but ... (Score:3, Informative)
It's about 1/5th the cost of the xserve raid but not nearly as flexible:
It's a different solution for different people. If you need reliability and performance and uptime, you get an xserve raid. If you need "good enough", you build one yourself. Same thing as getting a cheap dsl/cable router for $20 that "does the job", rather than getting a $$$ name brand router like a cisco or something, and a support contract, etc.
One will do the job most of the time, but when you absolutely gotta have the performance and reliability, if your job depends on it like the video editors you mentioned in your original post, the extra $4k for the xserve raid starts looking pretty good.