Sony develops new 185 TB storage tape
bbc.com
bbc.com
Currently LTO-6 2.5TB tapes are just useless. Whereas they are generally more reliable than an hard drive, for the same investment you can just buy more drives and make up for it, and you gain improved r/w speed and _random_ access, which cannot be overstated (tape is generally a PITA to use).
Not to mention that a single tape drive is only useful if you can backup your entire system with just one tape. When LTO6 came out, I already couldn't, making a tape robot necessary. Again, to make things into perspective, today you can just buy a cheap nas with 2 drives with hotplug, and you still saved 2k$ for the tape drive.
With a 185TB tape you can seriously think to backup an entire site to a couple of cassettes again. IF price is reasonable.
Not at all. Oracle and IBM sell data insurance. If you save data on their tapes, that data will be their for 100 years, or you get a large chunk of money. This is why corporations still use tapes, HDD's don't offer such protections.
Sony currently doesn't offer tape insurance. Which if they want this to succeed they likely will have too as this is the main attraction of tapes. Safe long term storage.
:.:.:
There are a few technical problems.
The biggest problem overall is write speed. A T1000C tape drive (5TB Oracle) can take about 4 hours to "Fill". This is with 16 parallel writes (tracks) taking place.
Having a flat 20x more data to write could mean times in the order of 72+ hour to preform a full back up (Assume 16 parallel writes, same speed as oracle), or LONGER!
I really want to see a data sheet on this new tech not hype.
Well tape drives don't offer insurance either... As you mentioned, Oracle and IBM do. If you have the cash you can insure anything you'd like, even hard drives.
HDD don't last 30+ years. Even if you take out a hard drive and set it on your desk, never touching it again chances are the data will be gone in 10-20 years due to bit rot (most likely at human survival-able temperatures). Tapes don't suffer from this (well they do, just at a slower rate due to larger bits).
That sounds exactly like a situation you would want insurance... Why insure the one that never goes bad?
Lets take life insurance for example, its very cheap on say 20 year olds, who are active, non-smokers. Because overall a 20 year old, active, non-smoker has a much lower chance of dropping dead over night then say a 80 year old active, non-smoker. Its pretty simple stats honestly.
Insuring a HDD for 20+ years is the equivalent of buying a 350 year life insurance policy for a human. Statically its pretty safe to say that policy will payout.
If you buy a policy that will statically pay out, it means the policy total needs to be collected before the event will likely take place, therefore result in a net profit for the provider.
:.:.:
Example
So you take out a 30 year protection policy on a 2TB hard drive, this policy will pay 50 million dollars. 80% of Hard Drives fail in 4 years, which means as an insurer you need to collect AT LEAST 80% of the policy pay out by year 4 of the devices life cycle. This means you'll expect payments of ~1.041 million a month (at least). For the insurer to break even on your policy over an infinite amount of hard drives (which they're likely using a finite model), so a lot more.
Now you take a 30 year tape drive, for 50 million dollars. 1% of drives fail after 30 years. This means you can collect the total 50 million from at least 100 different contracts, meaning within 30 years each policy will only collect ~5 millions dollars in payments (13k montly payments)
The former seems completely idiotic to both the insurance agency AND the business asking for it. I don't know what your point is.
:.:.:
TL;DR: For insurance to work (profitably for both parties) you can't have the certainty of the event occurring within the policies term.
I would love to see how you make this miracle happen, because it would make my life easier.
My numbers currently favor tape heavily. LTO-6 media is less than $25/TB, whereas spinny disks are $40/TB, and if those hard drives are going to be plugged in, add a minimum $10/TB for connecting them to something.
But most importantly, a tape robot costs orders of magnitude more. And if you don't have one, you either need multiple drives or fit the whole archive into one tape.
Now, if at home you can easily reach 2TB per-disk and have ~10TB easily, a nas with 8-16 disks in any work place will easily be too large for a single tape to be useful. A tape robot will be out of reach (in economical and logistical terms).
In these conditions, you really need a _huge_ system to make tape economically viable, but that's a very small market, whereas buying harddrives and cycling them will be the only logical choice.
Tape used to be different, where you could fit a week of incremental backups into one in the eighties. That made tape worthwhile. Then HD became more dense, while tape struggled.
I'm thinking of backing up on the order of 100TB a month, in which case a tape robot starts to make sense financially, but is a pain when it comes to software.
With hashbackup (http://www.hashbackup.com/ - not related in any way, just a happy user), we can squeeze roughly one year of history using a differential daily/weekly/monthly schedule.
It's faster than tape, and much easier to manage. If you account for electricity though, it's not much cheaper. But as we scale our storage, we can scale the other site as well (replacing hdds with higher capacity), which has allowed us to move from ~100TB to the current size without any change.
That being said, if I had a high density tape (as advertised), I would definitely choose tape. I don't think I need to explain the benefits of having multiple copies.
Anecdotally I currently backup about 100TB of data each month that get shipped off to a secure storage facility. I would love to be able to back up to HDDs/NAS and ship that out but the equipment cost is much higher and hard drives are not reliable for long term storage as the heads stick to the platters.
One thing I've done, without special software, when I've had to archive an entire system on a tape is:
1) Use "find" to make an index of all files to backup, splitting it up in approx. 2GB segments 2) Tar each segment in a separate session 3) Create a tar at the end of the tape with the index from step 1. The first write session of the tape contains a tar file with a README describing how to get the data (seek to the end of tape, back up one session, restore the index, then use that to position the tape to the session with the desired file). Also included was the shell scripts used to create the tape, along with a script to automate restoring a given list of files.
Just stick with "one-generation behind" unless you really need the space. LTO-5 tapes are $25 for 1.5TB of data. Plenty cheap if you actually use enough tapes to make the $2000 tape-drive worth the cost.
I remember about 20 years ago, it was quite reasonable to use tapes as backup even as a consumer.
If one generation behind would cost less than $500 I would think it would actually capture some of the consumer market.
Otherwise consumers have cloud and BD as backup and that's it.
Deleted comment
/Drops tin foil hat.
Really? I know lots of folks, within the industry and without, who use cloud backup services like Carbonite and Crashplan precisely because they are so much more convenient than maintaining a physical backup scheme, however simple. I veer in the other direction, because I trust USB HDDs in my physical possession more than I trust Crashplan or Carbonite not to screw up, but the argument from convenience is not an unreasonable one, and I'd say it certainly does count as the cloud "respond[ing] to popular demand".
It's not mutually incompatible: I trust both combined more than I trust each one alone.
What I consider "cloud" is closer to google app engine.
I mean, it's not as though you can't buy that sort of equipment and media in the relatively low quantities you'll need for individual use; you can easily, if not cheaply, purchase a SATA LTO-5 drive [0] and tape cartridges for it [1] right now, online. Sure, you'll pay a premium price per item, which a large organization placing an order of commensurate size wouldn't pay -- but that's hardly unique to this class of products, and strikes me as awfully scanty basis on which to construct a conspiracy theory.
[0] http://www.amazon.com/Quantum-LTO-5-Height-Tabletop-Bundle/d...
[1] http://www.amazon.com/LTO5-Ultrium-1-5TB-3TB-Case/dp/B003KR4...
Ironically the lack of a consumer grade system is bad for business. A lot of companies have remote offices with crappy connections that could be kept in sync with discs in the mail.
As for synchronizing remote offices, why not BD-RE? I can't imagine there being all that many companies at the intersection between having enough remote offices to care about synchronization, having those offices be in places without reliable backhaul, and needing to keep so much data in sync that one BD-RE 16x drive per office, and a relative handful of discs, would be an impractically large expense.
As the tape began to wear out, it would 'stretch' and subtly change the frequency of the data stored on it, causing issues.
Anyone have any further details on this?
In the bad old days there were multi-platter disk drives that used e.g. the bottom surface for a servo track.
Tape drives are rated for how many times they can be written / read to from end-to-end. LTO drives are rated for 200+ end-to-end reads.
So an LTO6 tape (2.5TB) has to have 500TB of data read / written to it before it wears out.
Tapes themselves are pretty darn cheap. LTO-5 tapes cost $25 for 1.5TB (uncompressed). Multi-layer Blu-Ray is the only solution that is anywhere close to that cheap... but those are a pain-in-the-ass to use. (You'll need Thirty 50GB Blu-Rays to do the job of a single LTO-5 tape)
Also consider: At x8, BluRays are written at 36 MB/s. Tape Drives are read/written to at 140MB/s, making them much faster than BluRays.
The problem is that tape drives cost roughly $2000 each. So tapes only become economical when you're storing 25+ TB of data.
On no! My Tape Drives are written too fast that they're causing me issues :-). If only my hard drives were fast enough to keep the write buffer full!
The L in LTO is for linear: when the tape writes its traveling at over 100 inches per second. Compare to helical scan tape systems like VHS and DAT/DDS where the tape heads are moving very quickly past the tape but the tape itself is not really fast at all. So if the write buffer empties, the drive has to decelerate the tape, wind it back quickly, then accelerate it so it's at full speed when it's at the location to write a new block.
I might recommend DDS for the 3rd World ^_^. It's what I used until it fell too far behind hard disk sizes.
The latter wasn't available when I set my current home system (the limits of local disk back where firmly confirmed by this: http://www.ancell-ent.com/1715_Rex_Ave_127B_Joplin/images/), but LTO-4, which will hold 800 GB per tape before compression and which while perhaps obsolescent is still widely used (hits some nice price performance points), was and I think still is quite feasible, albeit costly. But then again, how much is your data and piece of mind worth to you?
HP 1760 drive is on the the slow end, but that's fine for consumers with much less severe time to backup and time to restore requirements. Newegg price is a bit above $2K, with some serious bargain shopping you should be able to get one for substantially less. Add at least one SAS adapter (I use two), I like the LSI 9211 line, and a -4i (4 internal ports, BTW, they work with SATA drives as well), current Newegg price $163, and a fast but small SAS drive for buffering, say a Seagate Cheetah or Savvio 15K drive of 146 or 300 GB size, Newegg has an anomalously priced Cheetah 300 GB at $168 right now, call it $200 plus or minus, and then of course media.
There you want to get serious about bargain hunting and bulk buying (5-20 tapes at a time), but Newegg's single quantity pricing for Fujifilm or HP tapes is $40 plus or minus 2, and fairly easy bargain shopping I just did on Amazon to buy a few tape found HP tapes at $24-25 each from high reputation merchants.
dragontamer at https://news.ycombinator.com/item?id=7699179 says LTO-5 tape is going for $25 each, which would match my experience with LTO-3 tapes costing about the same as LTO-4 when got my drive; I just use 9 tapes for Bacula incrementals and differentials for a while, then move them to the full pool.
Its possible to max out the 6GBps channel (ie: 480MB/s) with say... a $90 120GB Vertex-3. (http://www.amazon.com/OCZ-Vertex-2-5-Inch-Performance-VTX3-2...).
The M.2 SSDs and PCIe SSDs are faster of course, which is why Consumer PCIe SSDs are getting pushed out by LSI and others. There isn't much point to faster SSDs when you're limited by the 6Gbps SATA port.
For example: $300 for the 240GB Plextor M6e, with read speeds measured at ~667 MB/s: http://www.newegg.com/Product/Product.aspx?Item=N82E16820249...
So SSDs are actually both cheaper, and more performant than a 15k SAS buffering drive!
The TLC SSD (3-bits per cell) is showing minor issues after 600TB of writing, but is still a usable drive. The MLC (2-bits per cell) SSDs have shown no issues after 600TB.
In comparison, a tape-drive would have long-past its ~200 end-to-end reads rating. MLC SSDs wear out slower than a tape drive! (Please don't take this sentence too seriously. I know I'm using hyperbole)
The SSDs I've linked to are MLC (2-bits per cell) designs, and can be expected to have roughly the same endurance as the MLC SSDs that techreport is testing.
By far the most telling takeaway thus far is the
fact that all the drives have endured 600TB of
writes without dying. That's an awful lot of
data—well over 300GB per day for five years—and
far more than typical PC users are ever likely to
write to their drives.
That said, at least for my purposes, using one such SSD drive as a 5+ year "consumable" would work just fine, and the price for the Intel 335 is just fine.
(On the third hand, since all my system "disks" are LSI RAID 1 2 disk mirrors, I had all the infrastructure to add a 3rd drive, and I trusted Seagate enterprise SAS drives more than I trusted SSDs of that class 3 years ago (not that the latter were big and durable enough back then, unless perhaps you went to the $$$$$$ enterprise level of SSDs drives), I use the drive for some other backup systems, like sending my most important data to rsync.net using rsyncrypto, so I need some speedy disk to disk action. When I do a major system refresh in 2+ years I'll seriously consider SSDs, or if I have to replace my LTO-4 drive before then and need to seriously look at LTO-5 drives and media; I just passed the 2 LTO-4 tapes per monthly full backup threshold....)
SSDs just have advanced surprisingly far in these past couple of years. Be on the lookout for PCIe SSDs on your refresh! They might be ready by then.
I personally just keep around a couple drives and backup one to the other.
A relative of mine recently needed 3 dvds to backups absolutely everything. She tends to delete what she doesn't want to keep though.
Me personally, I have a few hundred gigs I'd like to never lose... 3 or 4 consumer grade blu-ray discs should do it...
I wonder how long it'll be until a flash drive can hold this much data? I can usually find any song I want to listen to on youtube, but I think it'd be interesting to just have every song locally.