IBM one-ups Sun with terabyte tape drive IBM has launched its 1 terabyte tape drive the day after Sun launched the T10000B, and its is 33 percent faster than the Sun product. The IBM System Storage TS1130 is the latest generation of the company's 3592 tape format. The previous TS1120 drive stored up to 700GB of raw data on a 3592 tape. The TS1130 has an I/O …
Who needs SSDs, ordinary disk arrays will already provide a better backup speed in a disk-to-disk setup, and disks are not $19k a pop either! And there's not the risk of losing a tape in transit (civil servants reading this please take note), you simply replicate between arrays instead with a nice, cheap SATA or FATA array at your DR site. No need for expensive dark fibre links anymore, lots of FC over IP solutions over ordinary WAN links around.
If you don't want to use expensive arrays and replication, there are several cheaper virtual tape array devices on the market that imitate tape devices so as to be compatible with existing backup software that expects a tape device. You simply backup very quickly to disk, then stream at leisure to another medium such as DVD, tape or more cheap disk.
I know there's still a lot of tape in use, but I do wonder how much longer it will be around. In the meantime, congrats to the IBM team for their exquisite timing! LMAO!
OK, what's the catch? The media obviously isn't going to be 10 bucks a throw, so you'd have to be doing a LOT of storage per tape drive to justify the cost (as the first poster said). Am I missing something? What's the special sauce that justifies a forty thousand dollar price tag for a tape drive?
Tapes are by far the best way to large scale backups. They are small, light, fast, and reliable. Hard drives are not very good for backing things up. They weigh too much and are not very trustworthy in large numbers.
Cost is not a factor. These drives are meant to be extremely reliable and very high quality. People don't want to deal with el' cheapo to save a few bucks.
You CAN'T have your backups in a networked array. They have to be separate from each other and the rest of the system. A direct lighting strike would take out all your primary storage and your backups. There is not a surge protector in the world that could help. Then how much do you loose? 100 million? Looks like you should have used tapes.
Tape for large scale backup - No way! For a single small or large site, DeDup/VTL and Tape for offsite storage. In a multi site situation - Look at DeDup/VTL and Sync/Async data replication - Minimize tape to a very small footprint - Tape operations and management is resource intensive.
That a HP Ultrium 690 drive, 400/800Gb drive at £1200 or an LTO-4 800/1600 drive at 3000USD seems to be a similar end result, same data rates, same tape style backup, for one hell of a lot less! Even if you bought a new one every year it would work cheaper, or even had an array of them for redundancy and speed...i just cant quite add this up.
@Matt
I too love the timing tho! :D
Would everyone thumping on about using disks to replace tape please kindly shut up? Yes, everybody thinking about tape backups knows very well that you can implement a RAID as a backup device. But you seem to be missing the point. When it comes to backup and archiving, companies are looking for something reliable, fast, and large; something that will fit all of their data in one relatively small package; something with which they can keep multiple backups (daily, monthly, quarterly, yearly, etc); and something they can take off-site.
Disk arrays are typically reliable, they are fast, and they can be large. The problem is that with RAID, you have ONE backup device, so there goes having multiple backups in any real sense. Oh, and you did notice that you can't take your RAID off-site, right?
Matt (above) phenomenally missed the point by suggesting using a RAID replicated to an off-site RAID. That may be acceptable for having something off-site, but it's not very good for multiple backups or portability. It's also not good for archiving. Oh, and what would you do if/when your network access is severed and won't be repaired for multiple days? I guess you'll just have to hope nothing happens...
There's a reason companies like really big tapes. They can fit their entire backup on one tape and store that one tape off-site. They can even take multiple backups and store them in multiple off-site locations. Also, last time I checked, tapes were much more forgiving then hard drive arrays when it comes to transportation and dropping the cartridge.
Long ago, when external hard drives did not have much in the way of capacity, tape drive backups made a lot of sense. These days, with pocket-sized 2.5 external drives at 500GB sizes, the short-comings of tape drives far out way their merits. A few of these cons are: RIDICULOUS high cost/GB (relative to hard disk), incompatibility issues even within the same product line, no real random access, and the fact that proprietary software may be required for backups. With a self-powered USB external drive, just plug it in and back up your data. Next, regardless of median, take it off-site and store in a secure location (just in case your equipment is stolen, the building burns down, or the like). Oh, and having more than one for incrementals and such also helps. I haven't recommended any of my clients within the last 4 years use tape drive backups... that is unless they have weak/no on-site security. I don't trust tapes any more than I trust any other backup median--eventually they WILL fail. Over time the tape's read/write head or the loading mechanism malfunctions and both the drive and tape are mangled. Regardless of the backup median, it's your plan B (in case of median failure) that matters.
I think more and more, IT personnel are realizing this and are less and less likely to chose the tapes as a backup median. The only reason IBM and others are selling these for so much is to compensate for the ever shrinking quantities sold. Now if the tapes were 10TB and around $200ea it might be worthwhile--1TB tapes aren't going to do much good for a corporate data center's needs.
Trouble with your disk-based solution is that while it's a shedload faster than the rusty sellotape, it's also a pretty poor solution for long term backups.
Your disks are fragile, need power and SAN/WAN/LAN links, cooling etc. On the other hand IBM's tapes can be offloaded (I'm assuming this drive is designed for some mega auto tape library a la the old 3494), and then all they need is a dry cave somewhere.
The power and cooling issue is become more of a problem with all these rabid "greens" busy painting IT as the tool of the "man" (or the devil, depending on whom you listen to). And there too I would suggest that tapes are obviously better than disks.
See also "spam"'s posting.
On the other hand - we've now got 1TB cheap SATA disks (with 1.5TB due rsn) - so why doesn't someone caddy these and use em in a disk version of a tape autoloader. Heck, I've seen 500MB 2.5" disks advertised for laptops, and these (I assume) have head parking anti-shock tech too (but I would think still need a caddy), so they'd be relatively resilient. Resulting autoloader solution would be pretty damn fast, consume a lot less power than a SAN, and I'd hope be about the same price as a tape library. Interesting line of thought perhaps...
Too many people on here saying "disk is better than tape" are thinking too small. If you have say 50TB of data, and you need to keep daily copies for a few months, monthly copies for a few years, and yearly copies for 27 years, how do you propose to store all THAT on disk?
Disk cannot be taken offsite, unless you put the disk offsite. However an offsite disk array can blow away all your data in a single shot if it dies/somebody does something bad etc. Whereas it would take thousands of mistakes to wipe thousands of tapes.
For those who are concerned about reliability - that's why these ENTERPRISE drives are so expensive - they are damn reliable. I use older 3592 type drives (an older relation to the one in this article) and they are exceptionally reliable, maybe 1 or 2 tape errors per year, and 1 drive error every 2-3 years, thats with 10 drives and writing about 5TB of data per night, and doing about 100-150 tape mounts every 24 hour period.
And tapes don't use electricity or need cooling when they're not in use.
So stop thinking about your own small environment and consider those with bigger requirements than yours, protecting data which really is critical (lives depend on it etc).
I get paid to sell enterprise backup software.
The first thing you should consider is not the Data itself but the application which uses the data. How important are these applications to your business? What compliance issues are there? From these you can work out the recovery time objective (how long does it take me to get the application data available) and the recovery point objective (how old is the application data that I can now access). Now these can be anywhere from seconds (ok milliseconds but that's not going to be solved by data protection software, you'll need storage replication and clustering) to a few days for RTO and years for RPO (litgation issues). Consider also that the older the saved data is, the longer the RTO becomes.
So you can define RPO and RTO for all your application data, now you can set about working out which technologies you can use to try to meet these objectives.
Yes, short term RTO and RPO can be best met with various disk based technologies, Continuous Data Protection, Snapshots, backup to disk, VTL, but as many have pointed out Disk arrays are cheap to buy, but they are expensive to run (power and cooling), they break (cheap disks break more often) and do i really want to keep multiple copies what is essentially the same data on another disk array?
Storage Technology companies love to sell you more disks, because they know that you'll have to renew them every 3-5 years and then they'll come up with all sorts of software options for which they'll charge you (probably) per Terabyte, and then there is the maintenance on top.. So they come up with VTL (let's treat a random access device as a serial device and pretend we're a tape library) De-Duplication software ( they nearly always replicate, aha! A second array). Don't get me wrong, there is a place for these (and other) disk based data protection technologies, it's as a first backup destination for nearly all data protection. I'd go as far as to say that de-duplicated disk is a pretty good place for most medium term (around 1 year) fileserver or NAS content. (bear in mind that compressed formats don't de-duplicate well), But I would never recommend that this is the only copy to be kept. Array based Snapshots and Snapshot replication are great for keeping multple backups of databases, but they soon mount up (2% of the size of the database 10 times a day is 100% of the database in 5 days and you've replicated the array anyway so double that again ) if you sell disk storage a great idea, otherwise medium-long term storage on Tape please, so I can move it physically between sites and it costs me very little to store it.
Then there is compliance. Most of these rules are pretty vague but what they boil down to is pretty much always the same. You must be able to provide a particular piece of data from a date sometime in the past and prove that this data has not been altered and possibly who had has access to this data from the time it was written. Some of these compliance issues can be met by backup software, many can only be met by an Archiving Solution specifcally designed for this purpose.
Usually these involve WORM technology and are often Tape based.
Tape in all it's different forms is going to be around for a long, long time. There is simply no alternative method of economically meeting those medium and long term RPOs and RTOs .
I'm assuming that a drive of this size is aimed at enterprise, and more than that, major programmes with primary data volumes measured in 100sTB.
Nice to see the last poster talking about restore. I can't agree with the chap(esse?) concerned with size, that tape is the only option for 50TB backups where lives depend on it. On the major programmes I've worked on which sound rather similar (have we met?), where I was responsible for 0.5PB of primary data, the only way we could meet our SLA around RTO/RPO was to make the primary systems as resilient as possible (multiple sites, active active across sites). Backups were taken as the data was too important to be lost in it's entirety but we ackowledged that if we ever got in a situation where we had to restore 50TB plus from backup we'd fail our SLA. The bottlenecks in backup infrastructure (not just talking tape drives here) don't match the growing data volumes or customer business requirements on major programmes.
P.S. replication isn't so bad so long as your network design reflects the importance of the services using it. Multiple resilient paths through different exchanges are a must - this is expensive, but we're talking major enterprise projects here aren't we?
Tape?
I'm back in the 70s!!!! wewt
True, RAID is a bad backup policy but a 1TB removable drive starts at $100
that's faster and is off site storage
(I'd bet most tapes for these backup systems are kept on the shelf next to the system itself, which is only really viable with asbestos tapes; held inside an asbestos building where everything is made from asbestos)
Dead bird because these guys already flogged the dead horse. (^o^)<!!
"(I'd bet most tapes for these backup systems are kept on the shelf next to the system itself, which is only really viable with asbestos tapes; held inside an asbestos building where everything is made from asbestos)"
I think you will find that any organisation that is spending $40k on a tape drive can find cash in the budget for of site, dedicated, vaulting services for the 1000's of tapes they will need to store.
And for all you disk fans this is ENTERPRISE class backup/archive hardware. Not SME. We have very different requirements.
"These days, with pocket-sized 2.5 external drives at 500GB sizes, the short-comings of tape drives far out way their merits... With a self-powered USB external drive, just plug it in and back up your data"
Last time I used an external 2.5" USB drive (WD Passport 80), the actual throughput was very low. In my own testing (capturing a VHS stream), it couldn't even capture uncompressed 640x480 at 30fps (about 1GB/min [17MB/sec]); it kept dropping frames. Even assuming that 2.5" drives have gotten a lot better since then, are you actually comparing them to a 160MB/sec (9GB/min) tape drive?!?
A had a client using WD Passports for their backups. The fastest backup throughput was about 430MB/min with Backup Exec doing software compression. The average was closer to 350-375MB/min. The average verify throughput (helped by 2:1 compression on most of the files) was around 700MB/min. Including their pathetically-slow Exchange server (which is far underpowered, and so does not back up or verify at anything approaching reasonable speeds), the full backup/verify (90GB) took 9.5 hours.
Now that client is using an LTO-3 drive and backing up about 119GB. The backup throughput ranges from about 500-800MB/min on the local system to 700-1700MB/min backing up two new servers (using Backup Exec's remote agents). The verify speed is a blistering 3200-5100MB/min (up to 7600MB/min [277MB in 2 seconds]). Including that same pathetically-slow Exchange server, the full backup/verify (119GB) of the original two servers plus two additional servers now takes 4.5 hours.
So yeah, if you have a relatively small amount of data, go ahead and use a 2.5" USB drive. Or if you don't mind your backup not being completed by the time you need to start work in the morning, go ahead and use it (of course, that defeats the purpose since your in-use files would not be backed up). But for anyone backing up any modest-to-large amount of data, external 2.5" hard drives just don't cut it.
"True, RAID is a bad backup policy but a 1TB removable drive starts at $100...that's faster and is off site storage"
1TB hard drive for $100? Off the back of a truck maybe...
I just love these disk-lovers who don't know what they're talking about. AC, have you actually SEEN the throughput of your beloved 1TB removable drive? Have you actually SEEN the throughput of tape drives (no, your old Travan drive doesn't count)? Modern tape drives have ACTUAL throughput up to 160MB/sec (9GB/min). I've actually seen throughput of 5-7GB/min on an LTO-3 drive. You're seriously delusional if you think an external USB drive will reach anything close to that. Hint: 480Mbit/sec = roughly 48MB/sec. In other words, your USB 2.0 drive has a maximum THEORETICAL throughput of about 2.8GB/min. Most 3.5" internal SATA drives don't even come close to reaching 160MB/sec.
So what were you saying about removable drives being faster than tape?
I've still to come across a high end Ultrium, LTO or AIT drive that works well after 5 years of daily work. The drive-heads gets worn out by all the tape dragged along them. Another problem is microscopic dust that gets into the drive and onto the tapes. Why not make it into an almost vacuum and use water-cooling instead of those noisy fans. To get 160MByte/s out of a drive you need lots of tracks and high speed, and it is very mechanical, failure is bound to happen.
"I've still to come across a high end Ultrium, LTO or AIT drive that works well after 5 years of daily work."
And how many hard disks have you seen working well after 5 years of constant use? Plus, the nice thing about tape is that if the drive dies, you still have your data stored on tapes, which are quite reliable.
To be fair, for $40k, you can get a set of drives that'll do far more than 160MB/s. My personal raid5 array will do that, and it cost less than $1k for drives and controller.
And you definitely can have sets of daily, weekly, monthly, etc backups on disk, just as easily as you can put it on tape. Backing up to disk is decent for lots of people, but it is more risky than tape, and lots of organizations don't need to reduce risk as completely as possible.
Plus you'll find that the people that are supposed to buy these things barely bat an eye at a $40k piece of kit. That's barely anything in the enterprise IT market.
1 TB Tape drive in a library can store more data than any disk array dedupe/VTL/JOB etc longer/safer and cheaper. If you are inept enough to buy 1 TB drive and put it in an external enclosure or in a server and only ever use 7 - 10 tapes then YOU ARE NOT GETTING YOUR MONEY'S WORTH. If you run a shop that has a couple of these puppies in a library that holds 300+ tapes you ARE GETTING YOU MONEY'S WORTH out of them. Do the math 300 X 1 TB = 300TB and thats not counting what you have in your offsite pools. 300 disks in a SAN would take 2 racks and (i'm going to take a wild guess here) 10 times the power and cooling and i can't send it off site! The larger your library and the more data you have the cost of tape becomes the lowest, most flexible, and safest backup medium. The AC that started his post with " get paid to sell enterprise backup software." could explain it better for all you SOHO and SMB know-it-alls who work at 1-800-GET-GEEK. So before posting my USB this and my 2.5 that learn before you make an ass of yourself on this site. Most of the readers here are at this site because A. They are part of an enterprise or B. Part of the industry. 1-800-GET-GEEK rent-a-techs feeling elite should go to slashdot and try to sound intelligent and see what happens. Most here probably just shake their heads or laugh. For the rest of you please go back to Toms or Anand where you belong.
Yours Truely;
A tired enterprise sys-admin after a long day
Joe, you're obviously so tired from running around with all your tapes you fogot the key fact that backup is only half the job, at some point you will have to undertake a recovery. Recovery from tape is slow (step 1, send for the tape from offsite storage; step 2, wait 24 hours for tape; step three, check it's the right tape; step four, send for the tape you actually wanted; step five, wait another 24 hours; step six, check it's the right tape; step seven, sequence through to the bit of data you need to recover....). Recovery from local or remote disk is much faster. And if I only want one file for my recovery, I don't have to sequence through the whole tape to find it, I just go straight to the file.
And as to long-term disk storage, disk-based archival arrays are the best way to meet the demands of modern data retention legislation like Sarbanes-Oxley. Such devices are designed to retain data for seven years (actually, as you could still be inputting new data in the seventh year since you first started, the actual lifespan could be fourteen years). If you use a tiered storage solution, as your data becomes older and potentially less likely to be read, you have rules that shift the data from expensive high-speed disk to slower but cheaper arrays, and then maybe out to WORM media when you are sure it's become static data.
Myabe you should give the AC backup sales guy a call and see if he can sort out a solution to help you get over your tiredness. In the meantime, I'm off to see what's new on Tom Pabst's site.
you make some good points though i think you need to talk to a BKUP/DR specialist. We tried to go down the tapeless path a couple of years ago and yes we had a call into AC backup guy who saved us a shit load of money by stopping us.
"And as to long-term disk storage, disk-based archival arrays are the best way to meet the demands of modern data retention legislation like Sarbanes-Oxley." If i were you i would talk to your SOX compliance mgmr and the AC storage guy if you think disk is the best choice.
AS for your tape 2step, try using a vaulting technology! NB enterprise vault for example tells you which tapes to bring back!! 7 steps? 24 hours to get a tape? wow, have you ever been burned.
I will agree with you to a point with disk. They are a wonderful solution for small RTO and RPO. Databases that are the storage behind your main apps obviously don't need months or years worth of backups on disk with RPO of hours or days.
Right? For compliance SOX or HIPAA you put it on something cheaper and send it to Iron Mountain for archival.
One way or another tape will be here as long as there is disk. Holographic memory if becomes a true enterprise reality could be both. But for now, you should really analyze the TCO of both disk verses tape backup solutions. You will find that we really need both. AC storage guy will tell you this unless you have more money than god like Microsoft and Google. As for whats on Tom's site most of it doesn't apply to me cause i use a Mac. LET THE FLAME WAR BEGIN
LOL. Now would someone point out to me that the site does do Mac compatible gear as well.
Peace :)
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
