Slashdot Mirror
Is LTO Tape On Its Way Out?
storagedude writes: With LTO media sales down by 50% in the last six years, is the end near for tape? With such a large installed base, it may not be imminent, but the time is coming when vendors will find it increasingly difficult to justify continued investment in tape technology, writes Henry Newman at Enterprise Storage Forum.
"If multiple vendors invest in a technology, it has a good chance of winning over the long haul," writes Newman, a long-time proponent of tape technology. "If multiple vendors have a technology they're not investing in, it will eventually lose over time. Of course, over time market requirements can change. It is these interactions that I fear that are playing out in the tape market."
"If multiple vendors invest in a technology, it has a good chance of winning over the long haul," writes Newman, a long-time proponent of tape technology. "If multiple vendors have a technology they're not investing in, it will eventually lose over time. Of course, over time market requirements can change. It is these interactions that I fear that are playing out in the tape market."
Magtape is the only viable medium for things which are actually "backups" as that term is understood in the professional IT arena. Every other possible medium for backups has faults which cripple it for one or more of the requirements which backups are required to fulfill -- primarily that's length of storage, but there are lot of other fun failure modes.
Sure, spinning magnetic storage, optical media, and flash drives each have some advantages for specific purposes.
But go pull the post-close EOY General Journal from 1996 off of one, I dare you.
And if you think that's an overly strict requirement, a) you're probably wrong, and b) I can come up with lots more that you won't.
My commercial backup guidelines are these:
You need it backed up on at least 4 pieces of media, of at least 3 different types, in at least 2 different cities, in at least 1 different state; bumping each of those numbers up by 1 is not unreasonable.
Only one backup can be on optical media; only one can be on spinning magnetic media, whether it's powered or not (this includes the cloud, and local external HDD backups, whether powered 24/7, alternating, or pulled and shelved).
Flash media is right out, as are SSDs.
I can pull 20 year old DC3000 tapes off my shelf and read them -- as long as I have a SCSI interface for the computer in question.
GNU tar is great that way.
LTO-9 goes to 25TB/cart, LTO-10 goes to 48TB.
Already announced.
And wouldn't it be interesting to know if that study was based on cartridge count or capacity?
Of *course* the cart count is going down, not *everyone's* data storage needs expand without bounds, and newer larger sizes imply some catch-up.
The T10kd which came out over a year ago, puts 8.5TB uncompressed on a tape (reusing the previous generation's media no less). So, basically the vendors are holding LTO back for unspecified reasons, probably in a vain attempt to recoup the last generations investment, or maybe to boost sales of their "enterprise" drives (T10kd, 3592).
I'm a fan of tape backup when managed responsibly, but there's a fallacy that goes with recommending tape for backups: because you can train semitechnical users to dutifully change tapes and carry them offsite (e.g. on a bank run to a safe deposit box), tape gets recommended for businesses who don't have dedicated IT. But the duty of of maintaining the backup gets delegated from the original trained user, and changing the tapes becomes the whole of the backup maintenance: no one actually verifies that the backup job is running properly. I've been on calls to clients who've diligently changes their tapes nightly, but the backup software has been crashed for months...
Tape media's greatest benefit is its long storage life. Providing you have the equipment to do it, you could read a tape created 25 years ago.
Tape media's greatest liability is its long storage life. Will you be able to find equipment to read it 25 years from now? If not, you have what we call write-only media.
I think that tape is going to disappear as a viable storage medium, at least in the small business sector. The equipment and media is expensive, and most small businesses don't have the resources to employ someone trained in proper media management.
The replacement is going to be offsite storage farms, whether from a third-party cloud provider, or farms owned by the company that needs the backups. As the per-byte cost of disk storage continually and rapidly falls and wide-area network (Internet) bandwidth increases, offsite/online backups are becoming more and more feasible. Data deduplication and image management software technologies mean that a company can have daily backups completely automated and available as far back as they want. Restoring a file or two from these backups is quick and easy. My company already supports several small businesses using this backup technology; as existing tape drives fail they are seldom being replaced with more tape hardware.
The downside of offsite/online backups is that bare-metal recovery of a failed system from those backups is still extremely time-consuming. Eventually the bandwidth will become available to make it viable; until then tape still seems to be the best option for bare-metal recovery.
Give me my freedom, and I'll take care of my own security, thank you.
When measuring the cost of backups, the cost of the media is often a small footnote. The cost of off-site storage can end up costing way more depending on how frequently they pick up, how long you store the tapes, and how frequently you need to do emergency restores.. Note that you left off the "time" component of the AWS cost structure -- the cost is *per month*. Still, AWS has some serious advantages over tape -- like the cost of robotic tape drives and the housing and maintenance costs that go along with them (if you have that sort of need). Plus, if you are a big enough customer, those 1 cent/GB/mo costs go down quite a bit.
the growth in cynicism and rebellion has not been without cause
It's the LTO drive itself that kills you, the drive and the SCSI card will set you back a couple $K to start.
It starts to make sense once the "backup bandwidth", the frequency of backups times the size of the archive, exceeds a couple TB/week, but it just makes no sense to spend $2K to keep a backup of $500 worth of drives, particularly in a situation where you just don't need to have the weekly tapes from two years ago and the last working tree is the only one worth keeping.
If you take the MTBF for a large hard drive, the averaged cost of failure per year might be like $50 per TB per year. (Total POOMA numbers, but it's probably the right order of magnitude.) HDDs have low fixed costs but a higher cost per gig, tapes have a much higher fixed cost but lower cost per gig, so wether one or the other is more economic depends on how much you're backing up, the level of reliability you need, so many different factors.
Don't blame me, I voted for Baltar.
Working in climate science I can tell you that tape for us isn't going anywhere. Our investment gets larger in it every year, at least monetarily and capacity wise. Several of our groups have growth curves that scale linearly with the output of the supercomputers, meaning our growth is almost exponential. Most of this data is static and doesn't really change once it's been produced, but it does need to be read from time to time. There's no other solution out there that takes little to no power to store, no cooling, and can keep the data for years with minimal loss of integrity. We have data that goes back to the 1940's that we have to keep almost forever, this historical data is hugely important in how we create the models and cannot be lost. So we have to have somewhere to store all that data for the long haul, LTO is the medium of choice because it's vendor agnostic, fast enough, cheap, and large enough to handle what we need it to handle.
"Some books contain the machinery required to create and sustain universes."-Tycho
Tape is great for archive purposes. It would be a shame if LTO dies.
The biggest threat to LTO is the lack of vendors developing drives. At this time only very few companies do R&D for tape devices. IBM, HP and Oracle are the only serious players.
Oracle develops only Enterprise class drives (T10K), not LTO. The market for those drives is quite small, which means the R&D cost needs to be recovered from a relatively low volume. That makes them bl**dy expensive.
HP develops LTO. They have to do R&D for LTO only, as they have no enterprise class drives. LTO is considered commodity so margins are too low to spend a lot on R&D. They will therefor struggle to be cost effective.
IBM develops both an enterprise class (Jaguar) and LTO tape drives. From a tech perspective, IBM are in the best position, as they can develop new technology for their enterprise drives, recoup cost in that segment, and then commoditize the technology in their LTO drives. Unfortunately IBM no longer wants to be a hardware company.
This does not bode well for tape technology in general. I've made a good living from it writing software for tape drives, but I guess all good things come to an end.
To Terminate, or not to Terminate, that's the question - SCSIROB
Didn't you get that papyrus-memo? Everyone wanted to require a Negative number on all the clay tablets. The scribes were looking forward to the overtime, the Donkey cart guys were in favor of shipping em from Helios mountain DR site, but Ceasar said no and settled on AD for Go-Forward only.
Time for a new Political party in the US (or two!) One is off the rails Other cant pony up a leader.
AWS perhaps even a bit cheaper. But any and all benefit is gone the instant they needed to restore anything substantial.
What counts as substantial? Pulling a number out of my ass, but 1/4 of all of your data?
To do this, you'd basically want to overnight to AWS some HDDs and restore your data to S3 while your HDDs are on the way to AWS. Charges:
Restore 12TB of data in a day from Glacier: ~$3600
Handling fees on 3 external HDDs: $240
Data loading: $100, or so. This is for eSATA.
Shipping: Figure, what, $75 or so?
So to retrieve 12TB from glacier in 3 business days, you're looking in the $4000 range. I've never actually done this, but I'm calculating 3 days as 1 day to get the drives to AWS, AWS loads your data that day, and then returns the drives to you via 2 day return shipping.
Note that this cost goes down significantly if you don't need the data so quickly. If you can restore the data from Glacer over a period of a week, the restore portion (line 1 above) will cost only $500 instead of $3600.
I'll grant you that all of the above are still pretty big numbers, but they certainly are not multiples of $10k.
All that being said, if you ever have to restore so much data from Glacier, either your business has suffered a serious disaster, in which case spending $4000 to get your data back is probably not a big deal, or you are not using Glacier as it was intended. I.e. you don't load today's backups into Glacier. You archive one of last year's backup snapshots into Glacier. Glacier is for data that you never want to see again, but may be forced to see again for some reason.
An example use of Glacier: a relative of mine is a retired doctor who kept his office space only for medical records storage in case he got sued at some point in the future over something and needed to get a record back. I told him he could save a ton of money if he got those records scanned and archived to the cloud. Together, we found a company that would scan all of the 35+ years worth of records, organize them, encrypt them, store them in Glacier, and shred the originals. I forget what the $ worked out to, but it was a huge savings. He'll probably never get sued and will probably never need those records again, but if he does, he can still get what he needs.
They don't grade fathers, but if your daughter's a stripper, you fucked up. --Chris Rock