Riddle me this: What grows as it shrinks? Answer: LTO tape The LTO (linear tape-open) organization has released a tape shipment report showing capacity shipped grew almost 18 per cent from 2014 to 2015, while unit shipments have been declining since 2008. The LTO organization basically looks after the LTO format and represents the interests of the LTO Program Technology Provider …
"...Eternal 5D data storage could record the history of humankind.."
http://www.southampton.ac.uk/news/2016/02/5d-data-storage-update.page
If this is to be believed then we could be onto a winner depending on how much that laser costs to buy "upfront"...
"Just under 20 million LTO cartridges were sent to customers last year. As a comparison let's note that WD and Seagate combined shipped more than 350 million disk drives in 2015; the tape cartridge market is less than 0.00567 per cent of the disk drive market in unit terms (ignoring Toshiba's contribution)"
Eh?? I make 20/350 = 0.057 = 5.7%
"The rate of decline is about 1,000 cartridges per year. If that continues, then by 2020 we'll be looking at 15,000 cartridges shipped a year, and just 10,000 by 2025."
It looks like somebody didn't notice the "k" on the Y axis. 19,000k is 19 million.
(An LTO tape has a physical volume of 0.231 litres. 15,000 of them would take 3460 litres, and fit in the back of two estate cars)
It's a pity these drives are so expensive - another £1000+ onto the cost of a server may be nothing to big enterprises but to smallish businesses it can be enough to dissuade them.
Even LTO-4 drives are still selling for a small fortune - if older but sufficient models like these were available at a more reasonable price many more businesses might be persuaded to use tape backup.
LTO tape drives must be expensive. Otherwise they wouldn't be as reliable. Cheaper means lower quality components being used.
Anyway although the drives are expensive the tapes are really cheap. And the transfer rate is faster than most hard disks and the new LTO-7 is faster than most SSDs too.
A new LTO-5 drive (1500GByte tapes native capacity) can be bought for 1200Euros and tapes price is around 20Euros.
A new LTO-6 drive (2500GByte tapes native capacity) can be bought for 1500-1600Euros and tapes price is around 45Euros.
A new LTO-7 drive (6000GByte tapes native capacity) can be bought for 2500Euros and tapes price is around 120Euros.
Used LTO-3 tape drives (400GBytes native capacity) on eBay sell for just 50 to 100Euros on average.
While used LTO-4 tape drives (800GBytes native capacity) for 150-250Euros on average.
First problem, is how to feed the LTO Drive.
As you know, it can only "stream" data unto tape effectively.
And the data is compressed on the fly.
So you need hight performance HDDs, or ram cache, or wahteveer scheme you have.
If you have a local tape and backup from drives, that will be your limit.
Then, in order for this to make sense, you have to change the tapes. And THIS is the problem. Somebody changing tapes is EXPENSIVE.
So, it is a no no if you are colocated.. only works on your own premises and iomplies going inside the CPD.. or having a robot.
When you start having robotic arms, the cost of the drive starts being irrelevant, the backup device starts getting really expensive.
So at that point you might reconsider and just use an storagepod, that is less expensive to operate than tapes, much more reliable (you know, yo CAN get your data for sure our of the system) and is also random access.
If you are a huge player, tape makes almost no sense, if you are small makes no sense at all, so it is medium enterprises that MIGHT be interested.
Frankly, I don't understand why people buy tapes...it ends up being more expensive.. and I said the same about big and expensive HDDs SANs vs Flash.
"When you start having robotic arms, the cost of the drive starts being irrelevant, the backup device starts getting really expensive."
Back in the old days, when DAT was king, you would get cartridge compatible drives, where the plastic cartridge would hang off the edge of the desk while the DAT tapes are stacked vertically.
It would start from the top, shift the cartridge up one slot and take the next tape till done, all automagically and cost a reasonable amount. They weren't always reliable (not suitable for a dozen tapes over a weekend).
Last LTO drive I saw had a number of slots for 6 tapes, and it would internally pick the next one. Worked well.
Holy crap was it expensive.
When your data exceeds one tape, things get stupid.
Wrong.
LTO tape drives don't need to write at sustained maximum speed. If you don't mind the backup taking a longer time you can use slow drives and NAS on network as the source.
Anyway just adding a couple consumer SATA hard disk drives in cheap RAID-0 configuation would be more than enough for up to LTO-6 at least.
LTO-7 to achieve maximum speed at least two hard disks with a 200MByte/s sustained transfer rate.
"Then, in order for this to make sense, you have to change the tapes. And THIS is the problem. Somebody changing tapes is EXPENSIVE."
Nonsense. If you need more tapes you buy a tape loader LTO drive and not a single unit. They are more expensive indeed but it is not a consumer type product.
LTO tape drives are speed adaptive anyway, if maximum sustained drive speed can't be achieve from the source the drive would just slow down.
Most of the points you make are correct.
I have been a user of tape since 556bpi reel to reel. I have written tapes and read them back 30 years later, myself!
The early streamers needed a lot of support to get useable performance. Modern LTO, not so much - the drives have huge buffers and slow down so they can keep streaming, even when eating lots of small files. SAS appears to place much lower demands on processors than SCSI, which also helps.
Tape changers are a thing, and so is redundancy - cost justifications depend on whether you are backing up porn or financial transactions, but yes, you can let the robot keep changing the tapes. But mind where you put that robot!
Archive is not backup, and backup is not archive. So having a changer do Grandfather/Father/Son and changing the tape set once a quarter is not unreasonable for backup. And a second changer, somewhere else in the same colo might be good. However, you still probably want a disk backup to cover the failures that are not total catastrophies if short term outages will cost serious money. It wont save you if the server catches fire, but it might help if the motherboard dies without warning.
You might also want to have an archive process where data goes to tape physically elsewhere, and the tapes go into storage immediately. If there is a zombie attack on the colo, at least you will have reasonably up to date data, and records that pre-date the hack-in, no matter how long the malware was hiding (disks and GFS get you nowhere with that).
So: no matter how big youa re, tape is best for long term.
Correct extrapolation is per LTO generation. Because:
No one buys a new drive for fun, there's the migration cost to start => in 2015 LTO3 is no longer economic, even with the costs already sunk
Cost per TB has been tracking downward with other storage classes, which has kept tape alive.
It is clear that LTO6 will be here for a few years, then LTO7. But without an LTO8, 9 etc etc, other storage classes will rapidly kill it within a couple of years, because the market tracks down on the LTO7 curve
And there won't be an 8 or 9, because the market size doesn't justify even $10M R&D to develop it.
By 2021-2022 it will be apparent that all current tape generations are uneconomic to produce for that reducing market volume. Then it is a quick death by 2023 -producers aren't charities, and it is better to shutter the factory.Customers will still want to buy the stuff theoretically but there will be none to buy, apart from last time buys.
"And there won't be an 8 or 9, because the market size doesn't justify even $10M R&D to develop it.
By 2021-2022 it will be apparent that all current tape generations are uneconomic to produce for that reducing market volume."
What nonsense are you babbling about? Are you unable to read or what?
LTO-8 , LTO-9 and LTO-10 have been planned and announced already.
What do you think data centers would do without LTO tape drives? Where would they put thousands if not millions of PetaBytes of data in the next years, uh? SSDs ? HDUs ? Optical media? There is nothing better and cheaper than LTO tape drives for reliable enterprise class data archiving and there won't be for many years.
And for anyone thinking "on the cloud" ... think twice. Cloud services use thousands of LTO tape drives in their data centers otherwise they wouldn't be able to archive a damn thing.
What do I think data centres would do without tape?
In the long term (like 2025), use slightly more expensive cold-stored HDD.
*some* cloudy stuff is archived on tape, but most (by data volume) is not. The info is all in the article. Tape is a minority interest.
8 and 9 don't exist (obvs). The roadmap exists. The tooling does not. hardware isn't made by leprechauns.
I wrote it above, exactly: there are customers who do want tape. But "want", and "are prepared to treble or quadruple their spend, to cost share the r and d with a declining industry" are different.
I expect you "want" it, and your business will be slightly spannered when tape dies. But I'm equally sure that if you ask your boss - hey, can you just sign off this budget for $30M , to fund this other companies r and d, they will laugh. Point is, *theres no one to split the bill with, there just aren't enough customers in your position, with deep enough pockets*
That's the way capitalism works
YOU CLEARLY HAVE NO CLUE WHAT YOU ARE TALKING ABOUT!
Yeah kill LTO and just use HDUs... Seriously? So why aren't Data Centers, Enterprises doing that right now? Why didn't they do it in the past years?
SIMPLE.
Hard Disk Drives and SSDs are far less reliable for any archive purposes than high-end enterprise class tapes like the LTO standard.
And archiving on LTO is a lot cheaper too. And it will still be cheaper in the next 5 or 10 years at least.
The upcoming Intel 3D XPoint might be able to deliver some reliable SSD drives (because NAND flash sucks big time) but it will still be a lot more expensive for many years than LTO tapes and it might never get as cheap as the tapes.
Your dreams are far from reality.
LTO-8 and LTO-9 and LTO-10 are going to happen regardless of what you think or expect which clearly is anything you have no clue about.
@ Justthefacts
"What do I think data centres would do without tape?
In the long term (like 2025), use slightly more expensive cold-stored HDD.
*some* cloudy stuff is archived on tape, but most (by data volume) is not. The info is all in the article. Tape is a minority interest.
8 and 9 don't exist (obvs). The roadmap exists. The tooling does not. hardware isn't made by leprechauns.
I wrote it above, exactly: there are customers who do want tape. But "want", and "are prepared to treble or quadruple their spend, to cost share the r and d with a declining industry" are different.
I expect you "want" it, and your business will be slightly spannered when tape dies. But I'm equally sure that if you ask your boss - hey, can you just sign off this budget for $30M , to fund this other companies r and d, they will laugh. Point is, *theres no one to split the bill with, there just aren't enough customers in your position, with deep enough pockets*
That's the way capitalism works"
All you have done is re-iterate "I see good technical reasons to go on buying this stuff"
I don't know your sector and have no reason to disbelieve you.
I do not dispute that it "would" still be cheaper in 10 years time* [for some customers]
I do dispute that anyone will (or is) investing 100M of their own money to make it for you. Low margins and a *shrinking* market. I've done loads of pitches, and I've just never seen one proposing to invest for a shrinking tech. Ever. They are all either - this market grows by 10% CAGR and will go on doing so forever; or this market is heading for the stars - 60% CAGR, disruptive play, blah
This is a pure numbers game % ROI. Take the emotion out.
You are the one emotionally driven here with your childish crusade against LTO tape archiving trying to tell that no one would be buying it and even LTO-8 won't happen.
You don't have a clue.
Data Center and Enterprise business sectors need LTO tape archiving. Without it costs would quadruple or worse.
You are babbling nonsense silly stats and numbers based on nothing trying to disprove reality.
It is not that the LTO tapes have just been released on the market now! It is its 7th generation.
Also IBM and Sun Oracle still have their own proprietary high-end tape archiving formats which are not LTO and they aren't going to stop releasing new higher capacity models in the coming years.
And you keep telling that no one is buying the drives and it is a declining market and other nonsense.
There is nothing better to replace tape drives technology. And there won't be anything for the next 10 years or more at least.
If a new better technology appears on the market surely the tapes tech might be in trouble and stop being developed BUT until then either tapes or waste of money and lack of reliability for anyone that is not doing his/her job properly in any Data Center or Enterprise.
@" Justthefacts
Re: Incorrect extrapolation - wrong commercial understanding
All you have done is re-iterate "I see good technical reasons to go on buying this stuff"
I don't know your sector and have no reason to disbelieve you.
I do not dispute that it "would" still be cheaper in 10 years time* [for some customers]
I do dispute that anyone will (or is) investing 100M of their own money to make it for you. Low margins and a *shrinking* market. I've done loads of pitches, and I've just never seen one proposing to invest for a shrinking tech. Ever. They are all either - this market grows by 10% CAGR and will go on doing so forever; or this market is heading for the stars - 60% CAGR, disruptive play, blah
This is a pure numbers game % ROI. Take the emotion out."
Please stop misquoting me: I did not say "no one is buying"
I specifically said "some people are buying, but declining"
The article presents a graph showing some people buying, but total unit shipments declining steadily since 2007. The article also made a point of noting that. This isn't a crusade, it's just reading a graph.
I'm going to have one more go at this, then give up:
"There is no better technology". So what? Declining market => other people are doing things differently. Whatever that is.
IThe only question for a tech manufacturer is - for a given company R&D opex, in 2017 which 10 of the 200 projects pitched by the separate.managers in their departmental road maps will be greenlighted. It will be the ones where they can show market predictions going UP. This is just everyday life in an actual engineering company that makes stuff, rather than services company that buys stuff.
The article is about LTO, not the entire tape market. If the tape market is in such dire straits, I'd expect to see fewer formats in future which would reduce the amount of development money spent in total. I don't believe that Oracle have said they've stopped T10K development or that IBM have said they've stopped Jaguar development so there's two companies who appear to be investing.
8 and 9 don't exist (obvs). The roadmap exists. The tooling does not. hardware isn't made by leprechauns.
As a leprechaun, I resent that statement ;-)
Hardware takes a LONG TIME to develop. I bet you a Guinness to a used clover leaf that work has already started on LTO8 at various leprechaun habitations across the planet.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
