Slashdot Mirror
Backup Tapes: Alive And Kicking
yootje writes "The Register runs an article about the future of backup tapes, which looks pretty good. Although some people say backup tapes are dead, tape systems continue to evolve. To prove that, The Register intoduces some new products that are about to come, like the SL8500."
Until optical media surpasses them in storage capacity, ease of use, and reliability, I don't see tape technology going anywhere. They serve a specific purpose and serve it well.
I don't need no instructions to know how to rock!!!!
Since prior stories have illuminated optical (laser) retention problems, tape does not seem as outdated as it once was. Tape's biggest problem now seems mainly cost. I had a 5GB Travan in a system and the per-tape cost was around $40. DVD blanks are around $1 for about the same amount of storage.
We tried backup to disk in house to see how it would behave - backing up big SQL Server clusters.
The problem that killed it for us is when you're transfering to an 80 gig drive over firewire, you completely hog the hell out of the system, making it all but unavailable during the meantime. I don't know of any way to "throttle" the backup, there's probably some obscure tweak though.
Tape transfer rates are comparitively slow, which leave plenty of room for the computer to carry on it's tasks. Sure it might take all night to do a full back up, but the servers available during that time.
I don't need no instructions to know how to rock!!!!
We use 4TB of SCSI Disks, and removable 250GB firewire drives for backup. Tape has let me down way to many times. Plus, I can restore from a catalog with in seconds. I would love to see tape do that.
A RAID-5 array with hot spares or a remote backup site is much more reliable and cost-effective.
BZZZZZT! I'm sorry, but thank you for playing.
What happens when the CEO deletes his stack of porn off the file server? Your RAID-5 isn't going to help you one damn bit. And maybe your company doesn't have the bandwidth to move the 100+GB of data on the fileserver to an offsite backup.
Backups don't just cover hardware failures. They cover people failures.
Also imagine trying to do disk drive rotation for off-site storage versus the same thing for tapes. I'd prefer tapes any day given the delicate nature of disk drives.
The Register intoduces some new products that are about to come,[...]
The problem with all of these endless new tape technologies is that after they come they (or their vendors) tend to become lethargic and lose interest in the whole process so that six months later they're trying to sell you yet another replacement technology.
That's fine for something like a computer that can run the same software each generation, but for tape devices the need to change media is like having to re-code your application in a new language every time you upgrade the computer. People don't want to do it.
Most customers want a backup media that will still be viable in at least seven years because of legal requirements. That can mean needing to be able to buy a drive that can read their tapes 5-12 years from now. How many of these new tape technologies will have that kind of staying power?
The standard 9-track 2400 foot open reel tape served the computer industry for about 30 years, providing a standard storage and interchange mechanism for pretty much every computer larger than a PC. The Internet has rendered the need for an interchange mechanism less critical, but the instability in the archival storage formats is now giving people serious headaches.
G.
Only need to backup 160GB (do not forsee that growing in 5 years). Gonna just buy two 160GB IDE HDD's & 2 firewire enclosures.
~$340 for both. Keep one plugged in for daily backup, keep the other in a safe place... swap them every month.
Pretty cheap, plenty fast, and won't take up much space!
They also forget to mention that you can't just disconnect a disk array, send it off site for 30 days, and expect to easily restore it when it comes back. With tapes, even if the OS it was originally backed up on is Windows, and the new OS was Linux, it will work (seemlessly if the backup software allows). The other thing that most people ignore besides the above mentioned sneaker quality is the larger cost associated with rack space, power, and cooling when using disk. I can stack a hell of a lot more TB/sq ft. with tapes than even some of the highest density hard drives, and I won't have to pay as much for power. Also disk systems produce considerably more heat than a tape library.
None of this really matters to small installations, but to enterprise installations these things are a lot more important.
I sell servers to small SOHO type businesses, doctors, lawyers, accountants, etc. There is never large amounts of data but it does exceed the CD-ROM limits and DVD are just to unreliable. It is too easy to burn a coaster and they have poor shelf life. And even at 9gb they are often too small to put all the data on one disk.
And getting the office receptionist(often the person who will do the job of managing the media) to swap disks is often asking too much. It has to fit on one tape/disk/whatever or it isn't going to get done.
Tape especially DAT drives give most bang for the buck.
Slashdot, home of supporters of free software, free music, and free speech.Except for Moderators that disagree with you.
I always wondered why they don't use off the shelf VHS tapes for data backup. You could probably build an inexpensive, yet reasonably reliable backup unit from the mechanism+record/playback heads of a low end VCR.
You said it yourself. "Reasonably reliable." For the vast majority of us in the business world, the whole reason that we make backups is because disks themselves are only "Reasonably reliable." I'm paying for "highly reliable" or greater. Without it, I'll take my chances on a nice RAID array with redundant error-checking controllers or something and not worry. But if I don't know that my backups are good, then they're almost worthless.
As others have mentioned elsewhere, this brings up the good point - test your backups! And your whole disaster recovery scenario for that matter. If you wouldn't bet the company on a test, make sure that you're not betting the company on the real thing.
You're special forces then? That's great! I just love your olympics!
Oh. Well then. Carry on!
In that case, I think we're on the same page. For under a TB, especially if it's not a lot of hosts to back up, I'd go with hard drives, too. It's hard to make an argument for tapes with backup sets that small.
Now, for 90PB, I think we'd both have a hard time finding a HD-based solution that would be anywhere near the price/performance of tape.
49 20 68 61 76 65 20 74 6F 6F 20 6D 75 63 68 20 66 72 65 65 20 74 69 6D 65 2E
Customer: I accidentally modified this file 2 days ago, can I get a backup copy?
You: Sorry, you're screwed.
Me: Yes, I'll have that restored for you as soon as possible. How can I contact you to notify you that it is finished?
We use RAID-5 and tape backup (which is off-site). The RAID covers disk failures; the tape backup covers user screw-ups and disaster recovery. And we've used both frequently enough to make them worth the money.
For one I think CERN expects to generate on the order of 4 petabytes of data per year in a year or two. I think other large particle colliders may generate the same amount of data. Other places that might generate the same amount of that are pharmaceutical and genetics/proteomics/biology related projects.
"When you sit with a nice girl for two hours, it seems like two minutes. When you sit on a hot stove for two minutes, it
I dunno about you but my DDS4 DAT system tells me it needs cleaning with an LED, too. DAT works great so long as you respect its limitations: you don't use the same tapes over and over for years; you archive to them for years.
You're living in the past.
A RAID-5 array with hot spares or a remote backup site is much more reliable and cost-effective.
Have you forgotton that some places use tape systems for archival storage as well? I suppose near-line is dead as well. Most of the companies that I have worked at use high-rez artwork. At an advertising agency, you are churning out gigs of files that may be used for 6 months tops, yet in 2 years someone will ask for job # 232343-xxx for god knows what reason. We use a 100 tape library for back ups and archives, and the archive is integrated with the asset management system, so the studio bosses can go to a web page, see that a job has been archived, add it to his cart, and restore it from tape with out much more than punching 2 buttons. They can also archive the inactive jobs by adding them to a archive cart and telling it to 'go'. Even a Terabyte capacity raid fills up pretty quickly when a high resolution art file is 1-2 gigabytes. Having a near line archive that is part of our backup solution is saving us big bux. Don't know what raids you use, but these LSI logic fiber channel raids aren't cheap.
Even the changing formats arent that bad. The jukebox lets us upgrade the 4 drives that it contains. Keep in mind that this is a small scale solution too, Can't imagine keeping a bunch of video data live either.
One last thing that we need to plan for at a real business that relies on the data that it stores is a disaster. Should the server closet all of a sudden fill up with water, smoke, fire, server eating cockroaches, be smashed by terrorist piloted airplanes, etc etc. we always have an offsite backup. Lets see your raid 5 recover from being melted into a blob of metal. Clients ask about these things too, in the review process, and it is important that a client's digital assets that it pays real money for are protected. These things are real, not paranoia. I have walked into a smoking server room a few times...
music lover since 1969
No, the point being exactly what I said previously: sometimes it's quite a bit cheaper, sometimes it's quite a bit more expensive. Cost of a hard drive and it's enclosure comes to under $100 per 120GB. Nobody wants to keep just one backup, regardless of how much data they have, so figure at least three. The drive itself can hold daily, weekly, monthy, and yearly snapshots for 95% of the businesses in America.
Real world example: local company I contract with has 20 employees, making them over twice the size of 90% of American businesses. Their central fileserver has 40GB on its main drive. The 120GB backup drives can hold a full backup and several years worth (remains to be seen how many) of incremental backups, that users can easily dig through if they want to find a file they may have had a few months ago (just add the backup drive to the Samba shares). Swap them every day and keep the other offsite.
In addition to being way cheaper than tapes, it's WAY more convenient for both users and the owner who's swapping drives.
Point being, sometimes (as in this is the case for most businesses) hard drives are a much cheaper alternative to tapes.
"Convictions are more dangerous enemies of truth than lies."
You're clearly uninformed - a 4U Nexan 'ATA-Beast' array with fiberchannel connect to a server holding 43 300G SATA drives RAID-5'd works out to about 11.72 TB - costs about $41,000 fully populated.
Or, if you want a bit cheaper - Promise VTrak 15100 (3U) with an Ultra-160 scsi interface and 15 400G SATA drives Raid-5'd is about 5.47 TB for about $11,000.
This is right in the same pricerange (in $/gig) as a giant spectra-logic 20000 tape changer with 200 AIT-3 tapes. about $85,000 for 31 TB of storage.
So, what it really comes down to is - 1) do you need 'offsite' backups? 2) how often do you have to do restores and find the right tapes, etc. 3) how quick do the restores have to be, and 4) how long do you need to keep the data.
For my company, we need backups for about 30 days, and haven't been able to muster off-site backups for a variety of logistical reasons even with tape. We have a guy who's full time job it is to do restores for our customers. We do do _some_ offsite backups but for the majority of our customers, we do not.
For us doing nightly/weekly backups to disk and saving for 30 days is about the same cost as doing it to tape, but we can do the needed restores much faster and without occasional physical manipulation. So, it looks like we are going to be changing to 'Disk to Disk' or D2D backups sometime soon.
Cost wise, the initial outlay is about the same, but for our business model, the speed of finding and making restores (including nightly incrementals) is really a win.
Free Online Dark Fantasy RPG - http://www.blackmud.com
The other part of this is that when tapes are going bad, you have some advance warning. You start seeing hard errors while writing to the media, and will have a chance to order new media before the tape is completely bad. With a hard drive, you don't know it's bad until it's too late, and is developing bad sectors.
Just because you had a QIC, Travan or iomega Ditto drive, and it was junk, doesn't mean all tape technology is unreliable.
I used up all my sick days, so I'm calling in dead.
Bah. We backup our RAID-5, and for good reason. [...] When we got hit by a hacker a few years ago, after we had expelled him from the system we just restored from tape. Show me your RAID-5 doing that.
You're missing the point. Instead of buying a large tape jukebox, buy a SECOND large raid-5 array that is about 5x larger than the first and then write backup images of the first one to the second. Ie weekly full dumps and nightly incrementals - then you can have backups from any time in the last several days, or from each week going back a month or so.
Depending on your mix of restores and the egos of the faculty involved ("Ignore those students and fix MY problem NOW!" - dont get me started about lack of practical computer knowlege some CS professors have) you might be able to more easily find, and more quickly restore your backups from disk images than you might from tape. And you can MUCH more easily verify-after-write your disk images than you can your tape images.
You'll find that a big raid array or two will cost in the same range as a big AIT-3 jukebox in $/TB of storage.
You LOOSE offsite backup though and the ability to buy more media so you can occasionally make long-term archives.
A medium sized RAID-5 Array with a smaller cheaper single-tape drive would address both issues and might cost less. It would also certainly have quicker restores.
Free Online Dark Fantasy RPG - http://www.blackmud.com
Well, I was thinking of plain old ATA as that remains cheapest for the moment. SATA is getting 4 disk controllers too these days, but they're still a bit on the expensive side ($50-$100) for this purpose.
AIT-3 is indeed twice the capacity, but the drives are in the 5K range as far as I can tell.
And none of the tape solutions we've discussed actually include a real lib (only a reader, unless you found a really cheap lib), while the IDE solutions I've suggested are all on-line (altho they would have disadvantages for offsite storage).
So, for 1 TB storage (2.6TB compressed):
$250 for computer, extra controller $50, 5 200GB disks $600 = $900.
AIT-2 drive $2000 + $50*20 tapes = $3K
AIT-3 drive $5000 + $50*10 tapes = $5.5K
2 TB storage:
$600 2 comps+controllers, $1000 10 disks = $1600
AIT-2 combo: $4K
AIT-3 combo: $6K
20 TB storage:
13 comps, 100 disks = $14K
AIT-2: $22K
AIT-3: $15K
Of course, at that size ide disks start having all sorts of other drawbacks, like power consumption, maintenance, etc. And AIT-3 will soon be cheaper for the installation.
The point isnt really that tape is always a bad solution. It's just it's getting to a point where ordinary disk storage is cheaper for even pretty large sites, and unless the tape companies do something really significant to their price/capacity they are going to get run over by consumer hardware driven capacity expansion as even their bread and butter customers storage needs are outpaced by consumer hardware. As the next strong phase of consumer storage growth is just starting up (driven by PVR type hardware and software) it will only get worse.