back to article LTO-7 has it taped, but when will 'bigger/faster' thinking hit the buffers?

Next year, the seventh-generation LTO tape format will be hitting the market. We are talking about 15TB cartridges. It sounds like an interesting media and it actually is if you need to store huge amounts of cold data. Cost/GB is still the best on the market for cold data, but I think there are other problems to be considered …

  1. Nick Ryan Silver badge

    What's the real capacity

    What's the real capacity? Without the lies and FUD of the "compressed" capacity?

    1. foxyshadis

      Re: What's the real capacity

      6TB.

      Funny that LTO raised its compression factor to 2.5 from 2 (last generation), when in my experience data is only becoming less compressible, not more.

      1. Simon Harris Silver badge

        Re: What's the real capacity

        And what is the example given?

        Photos.

        The size suggested (3-5MB) is generally about the size of a JPEG from, approximately, a 12MPixel image (depends, of course, on the quality settings chosen and complexity of the original image). These are already compressed and won't compress much more, so for this example we're back to the raw 6TB limit.

      2. Alan Brown Silver badge

        Re: What's the real capacity

        "data is only becoming less compressible"

        Yup.

        I have about 1PB backed up and the compression ratio has declined from 1.3 in LTO2 days to 1.05-1.1 now.

        More and more archives are precompressed and about the only compression gains are found in the mailspool and user /home directory areas - as these are tiny compared to the archives they don't affect things much overall.

        It's always funny to see the faces on salesdroids trying to sell us deduplication systems when they realise that 99.9% of the data on our drives won't benefit from that and we already know it.

      3. Anonymous Coward
        Anonymous Coward

        Re: What's the real capacity

        "We are talking about 15TB cartridges" - no, we are talking about 6TB cartridges.

        Otherwise, what's to stop the disk manufacturers branding 4TB drives as 10TB?

        1. DougS Silver badge

          Re: What's the real capacity

          If they start selling drives where the controller automatically compresses the data, you know it won't take long before marketing gets wind and does exactly that.

          1. toughluck

            Re: What's the real capacity

            In short, you can't expose compression to the user and still use the disk as a block device.

            Longer answer: Disks already employ compression internally. Read up on SSDs and how they handle various workloads and what happens to throughput when using SSDs to store pictures, music or video.

            It's not exposed to the user, though. In theory, you could use SSDs to store several times their stated capacity, for logs for instance, but this would open a can of worms, like how to address specific blocks or what filesystem to use with them?

            The SSD would effectively have to become a character device (much like a tape drive, in fact), and then it would be able to store data up to a compression-allowed maximum. You'd have to read it like tape, though, with accessing the inventory and probably metadata separately and then performing a "space" operation to get to the file and start streaming it.

            It would work, but it's a lot of hassle for no apparent gain. And probably some companies already have patents around this invention.

            1. Alan Brown Silver badge

              Re: What's the real capacity

              "Longer answer: Disks already employ compression internally"

              SOME disks. Generally at the cheap/nasty end of the scale.

    2. Voland's right hand Silver badge

      Re: What's the real capacity

      Same for throughput.

      Max uncompressed throughput is 300MB/s, not 600 which is still a HELL OF A LOT in an archiving job as you will be reading 300MB/s off the holding disk and pushing 300MB/s out to the tape at the same time while servicing continuous input from backup clients.

      For the reference, looking at the stats when pushing out to a virtual tape emulated on a NAS, the last full backup job on my home network pushed only 31MB/s and only about 8MB/s combined network client client throughput. If I move the virtual tapes from the NAS to local storage, I will probably push to ~ 60MB/s, 90MB/s at most.

      300MB/s is a hell of a lot for a tape job because you are necessarily writing out linearly in a single thread. 300MB/s are 2.4GB/s which is approaching the theoretical limit of what you can do on one dedicated core using legacy calls in userspace (without special acceleration) on a single socket or filehandle.

    3. PleebSmash

      Re: What's the real capacity

      "We are talking about 15TB cartridges."

      HAHHAHAHHAHAHAHHAHA.

      Please no, Reg

  2. Alan Brown Silver badge

    Tape speed is one issue, having spool drives to handle it is another (gotta be SSD, especially if you're feeding multiple drives or despooling jobs whilst spooling others)

    On the storage front:

    There are a number of HSM solutions out there. The prime idea behind them is to put a few hundred TB in front of the library so that anything accessed more than once a year is more than likely already on disk when hit a second time - and anything written to tape is held on disk for at least 12 months before evaporating from the frontside cache.

    The single biggest problem with this (apart from taking 30 seconds to 3 minutes to reach an archived file) is that PHBs tend to reason that disk arrays cost about the same. It's not even worthwhile considering HSM solutions unless you have at least a couple of Petabytes in coldstore as that's about the point that power consumption savings start to balance out the cost of the libraries and drives to go in them.

  3. Lusty

    Cost

    While tape may have the best $/GB for storage media, it doesn't have the best £/GB for backup data, or even close to it.

    Tape, by its very nature, means having a whole second copy of your data, so if I have 100GB data and change 1GB per day then I need 107GB of storage for my backups.

    Assuming I have a replica at DR, with disk based backup my backup data in this same scenario would only be 7GB on top of what I was storing anyway, making SAN the better $/GB for backup data.

    There are people who still need tape, but the majority no longer do when using appropriate backup schedules and retentions. Almost nobody I speak to understands retention unfortunately so I inevitably have to explain why keeping a backup for a year is a poor idea :(

    1. Anonymous Coward
      Anonymous Coward

      Re: Cost

      Replication != backup

      If you have a 100TB replicated data set, and you issue rm -rf * on the master, the replica will quickly follow. Replication plus snapshots gets you closer though.

      Replication != archival

      There are business and regulatory requirements that non-current data be archived.

      Also, block level (SAN) replication isn't suitable for all applications, in particular databases; if you want to avoid data loss you'll want to use replication of the transaction logs, not the underlying filesystem which holds the database.

      1. Lusty

        Re: Cost

        LOL AC your understanding would be spot on if you hadn't missed the recovery points available in that disk system. In reality if you delete everything on one of my solutions, you create a snapshot which is empty (but the changes require no new space) which I could then roll back on either primary or recovery SAN. You, sir, are one of the people I need to explain this stuff to on a daily basis.

        Backups and replication are not archival. Keeping a backup tape for a year is NOT compliant with any regulations. If you need to keep an archive, keep an archive which is a complete record of data, which a daily backup is not.

        Block level data replication is just fine for databases which can be quiesced. I never mentioned sync rep, you just filled that in with your excitement at correcting someone.

        1. the spectacularly refined chap

          Re: Cost

          LOL AC your understanding would be spot on if you hadn't missed the recovery points available in that disk system. In reality if you delete everything on one of my solutions, you create a snapshot which is empty (but the changes require no new space) which I could then roll back on either primary or recovery SAN. You, sir, are one of the people I need to explain this stuff to on a daily basis.

          The precise issue doesn't really matter though: it is still a single system vulnerable to a single upset. It is all too easy to get carried away reading a paper on e.g. ZFS and thinking that it answers all your problems at a stroke. Such solutions don't exist - look at how many reports you see online about entire ZFS pools getting bricked. Sure, you've got 1,000 snapshots in that pool. It is completely meaningless when they are all lost.

          Funnily enough, I have to warn against the fools advocating magic wand solutions on a daily basis...

          1. Lusty

            Re: Cost

            WTF? I said two systems in two locations! I would never advocate using a single disk system with no second copy, but having two disk systems in two locations each with their own copy of the data and their own snapshots is perfectly safe - please tell me what the danger is with this scenario.

            1. Peter2 Silver badge

              Re: Cost

              Hostile action. ie;being hacked, or an employee with administrative access writing everything off when they get fired.

              Happens quite a bit. Businesses have gone out of business because their systems have been writen off and the backups deleted along with the live systems and this has been covered on El Reg. While a bit BOFHish, this does make me laugh.

              Disaster Recovery plans and supposed to plan for DISASTERS. Who plans Disaster Recovery with a mindset of, "oh, what are the chances of that!" You plan with a mindset of "What are the chances of that?, what is the severity compared to likelyhood and how can I mitigate against it." Offsite tapes are a nice little mitigator against pretty much everything.

              1. Lusty

                Re: Cost

                Employees can't do anything when they are fired - you revoke their access before you tell them they are fired.

                I've never come across a hacking incident as you describe, but I guess in theory it could happen. But then, it's at least as likely that Iron Mountain could burn down when you need your tapes. Probably more likely for the majority of businesses who don't have hackers out to get them (so, everyone but Sony really).

                1. Peter2 Silver badge

                  Re: Cost

                  Employees can't do anything when they are fired - you revoke their access before you tell them they are fired.

                  In theory, yes. In practice, have you seen some of the tales of vengefull admins leaving on el reg that make the BOFH look quite tame...? They don't even need computer access if they just come in and carry the equipment away. (Which, again has happened on articles covered on El Reg).

                  I've never come across a hacking incident as you describe, but I guess in theory it could happen.

                  For your reading enjoyment.

                  http://www.theregister.co.uk/2014/06/18/code_spaces_destroyed/

                  Doesn't look Sony sized. You can find plenty of other examples on this site if your interested in looking.

                  But then, it's at least as likely that Iron Mountain could burn down when you need your tapes.

                  With respect, I think your missing the point. My offsite tape storage could burn down at the same time as the office (rendering me unable to get to the fireproof tape safe on site with a burnt down building on top) and i'd still have tapes at my home, older archive tapes stored at other offices and by the firms owner. Because a Disaster Recovery plan is supposed to consider and mitigate against worst possible cases. (Disaster Recovery only being needed in a disaster and all..!)

                  What would the severity and expense be of you losing both sets of your online storage due to a virus or being hacked? If it's under the cost of backups then it might not be worth doing. For most businesses though the cost of losing everything is extreme and the cost of doing backups is tiny in comparison.

                  1. Lusty

                    Re: Cost

                    "In theory, yes. In practice,"

                    Yes, and in practice I've seen more tapes with gibberish filling them because the admin has no clue about backup than I have hackers wiping 2 entire SAN systems due to poor security practices. In my experience, those clinging to tape do so because they are scared after a bad experience (i.e. they F'd up) rather than because they understand risk and requirement.

                    If you have tapes in your house then you'd be the first one I'd sack for breaking of procedures. It's wholly inappropriate to have copies of corporate data in your own home as there is no way to revoke access if you're fired or some other HR procedure is enacted.

                    You mention older archives (Archives are not backup and vice versa), with no mention of the business value in recovering out of date information. A database from 6 months ago is worthless in most circumstances - a database from the day before yesterday is worthless in most stock control systems!

                    You mention a virus as if it's a risk I didn't deal with, snapshots will give identical protection in a virus outbreak to tape, only my recovery time is under a second and yours could be weeks.

                    As I said, write down your actual scenarios and work out what will and will not cover you. I have yet to find something other than mythical hacking incidents which would require a tape. Even then, I wouldn't need tape to have a read only copy.

                    1. DanielN

                      Re: Cost

                      "You mention a virus as if it's a risk I didn't deal with, snapshots will give identical protection in a virus outbreak to tape, only my recovery time is under a second and yours could be weeks."

                      Until the hostile software runs on the storage computer. Then careers end on the evening news, and maybe people go to prison.

                      "It's wholly inappropriate to have copies of corporate data in your own home as there is no way to revoke access if you're fired or some other HR procedure is enacted."

                      Competent backup systems are encrypted. The encryption key goes to trustees, and the tapes go to custodians. The custodians need only be reliable, not subversion resistant or trustworthy.

                      1. Lusty

                        Re: Cost

                        Competent systems don't include taking tapes home. I'd love to see an example of a virus which can automatically infect storage systems on two separate sites. I assume you've a link to information about one?

                        It's infinitely more likely, however, that your Windows system (which has many, many documented viruses) holding the encryption keys you mentioned will get a virus which destroys those keys, making all your tapes effectively blank...

                        1. Anonymous Coward
                          Anonymous Coward

                          Re: Cost

                          Sleepers. They wait for some time so as to get into whatever recovery systems you have THEN strike.

                2. Anonymous Coward
                  Anonymous Coward

                  Re: Cost

                  Something like that, Iron Mountain equivalent did burn dow, in South America and actually may have been 'enemy action.'

          2. Alan Brown Silver badge

            Re: Cost

            "It is all too easy to get carried away reading a paper on e.g. ZFS and thinking that it answers all your problems at a stroke."

            ZFS answers several problems, which is why I'm in the process of installing a 400TB system at the moment, but it's not a backup or archive solution and never claims to be. Snapshots are for when someone fatfingers a couple of file deletions and that's about it.

    2. Steven Jones

      Re: Cost

      I have news for you. All robust backup regimes for commercial IT systems require multi-generational copies. Yes, you can look at using de-dup and incremental techniques, but if you don't have more than one completely independent copy, you are playing with fire. Backup also provides protection against corruption, whether hardware or software induced.

      As far as multiple online copies and redundancy is concerned, then major mission-critical systems will have those to, but that's largely to address availability and does not substitute for the recovery of last resort that true backup requires.

      It should also be borne in mind that online disk copies require power, and in a modern data centre, power and heat are serious issues. There is always talk of disks that power-down when not being used but frequent power downs and restarts are not recommended.

      And finally tapes are physically robust and easily transported off-site. I would not recommend doing that with disk drives (especially the sort of high-capacity ones suited to backup). They aren't designed for frequent handling. Yes, you can do offsite backups using disk storage but the demands on networks can be immense. Not to much of an issue for incrementals, but heave help you if you need to recover a 100TB database over a network.

      Tapes aren't suited for many loads, but they have a niche best described as write once and read virtually never.

      1. Danny 14 Silver badge

        Re: Cost

        I didn't read it as replica, more use a san as a storage for your backups. I have a "cheap" pseudo SAN in the form of a dell 3220 that is full of nearline drives. It is carved up into a couple of 12Tb RAID 6 arrays and host my backups (not archive, they are stored on paired BR and put away in the safe). The array was 4k (refurbished array, new drives) and blu rays are cheap. Now we aren't a powerhouse of data but the convenience and simplicity is helpful. People regularly lose files and pupils are forever deleting their own things so the speed of restoration is good too.

        Scale up and of course this isn't a good idea (we are only a few TB in data all told so this solution is small scale).

        1. Alan Brown Silver badge

          Re: Cost

          "The array was 4k (refurbished array, new drives) and blu rays are cheap"

          Tell me how much it would cost (and how well it would work) for my "small" system (which is 50TB), let along the larger ones (which start at 200TB and wind up from there)

          I'll stick with tape, thanks. The labour costs of having someone swapping out bluray are enough to wipe out any hardware savings you might make.

        2. Alan Brown Silver badge

          Re: Cost

          "I didn't read it as replica, more use a san as a storage for your backups."

          "Online/nearline backups" are susceptable to rm -rf events. Several ISPs have found that out the hard way when script kiddies have trashed systems _and_ their backups.

      2. Lusty

        Re: Cost

        @Steve Jones, I never said I didn't have multiple copies in multiple locations with multiple recovery points. What I actually said was there's no need for tape in a solution with replicated DR SAN, and there isn't for all the reasons you mention.

        Online disks do need power, but the DR copy is using that power whether you put tapes next to them or not.

        Modern data backup is wholly different to traditional backup. If you lose a RAID set you'd be in DR regardless due to the number of probable affected systems - a LUN in VMware may have 40 VMs on it, so I would fail to DR in that scenario rather than recover a backup to the broken SAN which would take many hours longer than the recovery time to a well configured and tested DR site. We're not backing up one server at a time any more, and if you take the time to look at the various recovery scenarios you'll realise SAN based backup fits them all nicely - the NetApp brainwashing, sorry marketing, explains this quite well if you need it spoon fed but just making a list of when and why you'd use a copy of the data will get you there pretty quickly on your own. When you do this, you'll realise that a failure in the extremely redundant and highly available SAN would be the only time you'd ever need a second copy elsewhere - and this is a DR scenario because your extremely redundant and highly available shared storage is broken. Every other scenario involves getting data from a recovery point, which can live on the primary storage without problem. The main blocker to this, is that people keep too many recovery points, thinking they are making archives.

  4. toughluck

    If you don’t sustain the throughput, the tape drive won’t have enough data to write and would have to stop/pause, fill the data buffer and reposition itself before restarting. Each one of these stops takes several seconds and it will heavily impact performance or real throughput.

    This is a well-known problem, which has already been seen and addressed in the past.

    Yep. LTO for instance, throttled back by half as early as its first generation. Improved since, and then all the way down to 1/16th of the speed with LTO-4. I've tested LTO-6 with various speeds and it can throttle down to 3 MB/s with no shoeshining.

    Taking the example of a picture database, the example given (pictures) is really bad. These days, a 3-5 MB picture file is metadata, a thumbnail, if you will.

    It's a thumbnail of the actual picture, which will be hundreds of megabytes large. A 120 megapixel 16 bit raw file is going to be 240 megabytes in size, readily compressible, but certainly not down by 2.5×. A TIFF file of that picture is going to be about 720 megabytes, but more readily compressible (indeed, 2.5× compression can be achieved). In the end, however, you're storing perhaps 150 MB of the raw file, maybe 300 MB of the largest TIFF, plus a number of resizes (which, due to Shannon-Nyquist sampling theorem, are less compressible since size reduction has compressed them already).

    All in all, you're probably in the ballpark of ~1 GB per picture (natively). That means a library of just 6000 pictures. A reasonably smart archive is going to stage the popular files on a higher tier, and will make more than one copy. Once the file is accessed more than a set number of times, it's going to drop that copy, will not reclaim space, and then make another copy to prevent tape wear.

    That's for pictures. If you consider movies, in a year or two, an average single movie master is going to be:

    7200 seconds × 120 fps × 3 images (stereo+center) × 8K×4K (8192×4320) × 48 bits (4:4:4 in 12 bits) + 22.2 audio in 24 bit 192 kHz

    28 TB for video and 90 GB for audio (funny how almost inconsequential it is) before compression. That's the thing that's going to be archived on these tapes (or, more likely, on Oracle's T10000E or IBM's TS1160).

    I can't even call this an article. At best, it's an advertorial for SpectraLogic.

    1. foxyshadis

      In a year or two? Maybe in a decade or two.

      Who would store a single movie on 2,000 tapes?

      I have to point out that those numbers as realistic as LTO's advertised 2.5x compression ratio. 120 fps? Even 48 fps has only been in a couple of films, 24 is still standard and all of the old 60i except sports have moved to it. 3 images? 3D is niche despite years of effort, being phased out, and that's still only 2 images, a left frame+difference (easier to compress). 8K is years, maybe decades from penetrating the studio market; they're still grappling with 4K right now. And typical movie length is 1.5 hours, not 2 hours. I don't work with audio as much, so I can't say how real-world 22.2@192 is.

      And while you can blithely say "before compression", every recording format is compressed these days. Some shoot straight to pro-H.264 (444@12 bits), some use proprietary codecs, but there are none that aren't compressed in some way.

      Real world data rates of current RED cameras at 4K are about 1.5Gbps, and that's likely to only slowly increase, not massively jump. Even for a 2 hour movie, that's 1.3PB, not 28PB. That's still huge, it's an order of magnitude less, but even at a mere 100 tapes, who's going to store a movie that way?

      More likely it's going to be recompressed to fit into a few TB and left at that, so one tape or one HDD will suffice.

      1. toughluck

        Re: In a year or two? Maybe in a decade or two.

        Thanks for spotting these and for correcting me.

        I made a mistake in my calculations, off by one order of magnitude. 28 TB didn't seem so small before lunch.

        It should be 500 TB, not 28 TB. Not 2000 tapes, though, just 84 LTO-7 (less for T10000 and TS1150, even less for next generation).

        If anything, I gave a ceiling of what you might reasonably expect as the absolute maximum that's going to be needed.

        In the coming years, more and more content will be shot on high frame rates and 120 fps is definitely the maximum above which nobody is going to care. 60 fps is going to be the sweet spot, maybe even as low as 48 fps. Nobody's going back to 24 fps, though.

        8K -- television set makers are demanding that and there's really no point in making movies 4K-only. I don't see the point of 4K, let alone 8K, other than the fact that it will make 1080p sets cheaper in the next few years.

        Storing just the left frame and difference is not effective, since it doesn't make the resulting stream easier to compress -- the shift between the frames is not the same in all planes, the end result would be probably harder to compress or you'd expect horrific artifacts. If anything, the two frames are stored side-by-side.

        You're right about storing just two frames, though. After all, a stereoscopic camera doesn't record three images, just two and must use either one or the other for 2D. So, 334 TB.

        90 minutes, 120 minutes. If you shoot state of the art, you're going to make it longer rather than shorter, but it doesn't make an order of magnitude difference.

        If my calculations were off, let's rehash them:

        5400 seconds × 60 fps × 2 images × 8K×4K × 48 bits = 125 TiB*

        *) 138 TB, I made the mistake of assuming binary prefixes previously (so an LTO-7 cartridge has just 5.46 TiB).

        Four times less, so just 23 LTO-7 tapes.

        That's just for the master. The full movie as-shot will have many retakes, so probably ten times that number, but still something that's more suited for tape than disk (although nobody's going to store the whole movie forever, unless it becomes a cult classic and will pay for its storage, so the studios will move back to reclaiming the stock for silver, so to speak).

        And indeed, a lot of compression. You're right that cameras and storage can't handle the throughputs and raw capacity required at 95 Gbps, so we're going to be getting there very slowly. But yes, a handful of tapes (or disks) for the resulting whole movie is about right. Doesn't mean that a studio won't need several libraries to store all those movies.

        1. DougS Silver badge

          Re: In a year or two? Maybe in a decade or two.

          You just ignored all the previous posters corrections and blithely continued with your mental masturbation about shooting 8K movies at 120 fps "in a year or two". If you want to live in a fantasy world, that's fine, but don't try to fool others into believing your bullshit. The small percentage of theaters that are now equipped with 4K digital projectors are not going to upgrade again so soon, so you can forget any ideas about 8K movies for at least a decade. While the UHD standard includes 120 fps, that's strictly for sports, no one is going to shoot a movie faster than 48 fps - and 24 fps will remain the norm for many years to come. It is questionable to many whether 48 fps is even an improvement!

          And studios won't need "several libraries" to store these movies no matter how big they are, because they don't need all the footage to be online for movies that are completed. They'll be in a vault somewhere, probably the same one under Kansas where all the master prints of films are stored. The ones being actively worked on will be on disk arrays, not tape, for obvious reasons.

          1. toughluck

            Re: In a year or two? Maybe in a decade or two.

            Yes and no. You'll see television set makers market their 8K sets and will demand content from movie studios. Cinemas no longer drive all releases, TVs are starting to take over.

            As for not needing all the footage -- when they want to release a director's cut, then gold director's cut, then platinum director's cut, then the ultimate director's cut, and finally the final ultimate director's cut (until the next movie in the series is released, at which point we start again), they will want all the original footage, thank you very much.

  5. Ben Liddicott

    Tape is on it's way out.

    * Tape storage halves in price per GB every 30 months

    * HDD storage halves in price per GB every 18 months.

    So at what point will HDD take over from tape?

    It already has in most SMEs - convenience is superior. Use a USB3.0 SATA dock (£50), and cheapo HDDs as "cartridges" - £30 per TB.

    At these prices snapshotting is a fine solution for SMEs with no de-duping needed, and full random-access capability.

    1. toughluck

      Re: Tape is on it's way out.

      Half price per GB every 18 months? Utter and complete bollocks.

      I bought five 1 TB drives in 2010 (exactly five years ago) and paid £35 a pop (cheapest WD Green jobs and one Samsung).

      Going by your estimations, the price today should be 9.38% of that £35. So I should be able to buy a multi-terabyte drive for £3.30 per terabyte. Nowhere to be seen.

      Ah, but there was a flood in Thailand in 2010, you may say. So let's take 2011 as the baseline. Oddly enough, I bought one 1 TB drive in 2011. The same WD Green and I paid the same £35 for it.

      By your estimations, price per TB should be 15% of the price on 2011. £5.25. I'd like twenty terabytes of that storage, but strangely, nobody's willing to sell at that price.

      The best I can find is a Seagate 8 TB archive drive (£22 per TB), or a Toshiba P300 3 TB drive (£24.50 per TB).

      An LTO-6 tape is £20 bought single (£8 per TB), probably half that if bought in bulk (so £4 per TB). LTO-7 is not going to be markedly more expensive, and even then, the extra capacity will more than make up for it.

      1. foxyshadis

        Re: Tape is on it's way out.

        If anything, that £25/TB price has been mocking us for years now; bought a 4TB for £104 over a year ago. Capacities are rising again, but prices are dropping much more slowly.

      2. Warm Braw Silver badge

        Re: Tape is on it's way out.

        Unless you have really large amounts of data to backup, the cost of the tape drive (say £1500) is quite significant - it would buy you roughly 30 2TB disks - and that's assuming you already have a system with the appropriate I/O bandwidth. While tape is undoubtedly still cost-effective for people who have huge amounts of data to preserve, it has been largely pushed out of the small business backup market.

        1. Joerg

          Re: Tape is on it's way out.

          "Unless you have really large amounts of data to backup, the cost of the tape drive (say £1500) is quite significant - it would buy you roughly 30 2TB disks - and that's assuming you already have a system with the appropriate I/O bandwidth. While tape is undoubtedly still cost-effective for people who have huge amounts of data to preserve, it has been largely pushed out of the small business backup market."

          If the small business market and small offices are full of IT people and managers with no clue of what they are doing and wasting money on the whole cloud backup services nonsense and SSD NAND backup then of course they won't buy LTO and do some proper archiving of all their data.

          Those people should just be fired because they put data at risk.

          LTO is the best way to backup data.

        2. Alan Brown Silver badge

          Re: Tape is on it's way out.

          "the cost of the tape drive (say £1500) is quite significant "

          Yes it is. More like £4000 a pop in libraries, plus the cost of the robot on top.

          "it would buy you roughly 30 2TB disks"

          Do you plan to leave them online? If not, do you have a decent fire safe?

          What happens when you need more storage as your backup sets increase? I can go and buy more LTO6 tapes for £14-18+VAT whilst 2Tb drives start at £55+VAT and go up to £200 depending on the application (would you really entrust your backups to the cheapest drives on the market?)

          Ditto when a drive goes bad. And what happens when (as inevitably happens), someone drops it twixt the drive rack and the data safe?

        3. This post has been deleted by its author

      3. Ben Liddicott

        Re: Tape is on it's way out.

        Fair enough. HDD prices didn't drop at that rate between 2010 and 2015. Excl. VAT I can find about £20/TB. Perhaps they've reached some sort of limit. That makes tape about six times cheaper per TB.

        http://www.jcmit.com/disk2015.htm

        Disk has much lower fixed costs though.

        1. DougS Silver badge

          Re: Tape is on it's way out.

          Disks make sense in many cases for daily backup needs, but for archive you'll always want tape.

          Sure, you could put a 2.5" drive in a LTO carrier case so after replacement of the LTO drives with SATA+P connectors it could work for offsite archive. Tape would still win on throughput and durability, however, so overcoming those disadvantages and the long history of tape would require hard drives to be significantly cheaper per TB, not merely reach price parity!

        2. toughluck

          Re: Disk has much lower fixed costs though.

          Cheapest consumer disks maybe. Enterprise disks, no.

    2. Joerg

      Re: Tape is on it's way out.

      You have no clue what you are talking about. Really.

    3. Alan Brown Silver badge

      Re: Tape is on it's way out.

      "* HDD storage halves in price per GB every 18 months."

      That "rule" hasn't been valid for at least a decade.

      Likewise the ones about computer speeds. They haven't gotten much faster in the last 7 years. All the development has gone first into parallel CPUs (remember when the received wisdom was that more than 4 CPUs was a waste of time?), and then reducing power consumption for the same speed.

      Memory speed has increased for sequential access but not for actual random access. Actual access latency isn't much different than it was 15 years ago (about half what it was, not the orders of magnitude needed to go with silly-fast clock speeds) - and ram density seems to be close to tapped out too, given that we're not dropping 1TB in a desktop system yet.

      Bang-per-buck has kept up with Moore's Law but the difference has been that $N buys a 2 CPU system giving X FLOPS, not a 1 CPU system giving 2*X FLOPS.

      Disks have gone about as far as they can go. HAMR may not hit the market in time to stave off 4Tb flash drives at comparable pricing - and the access time of drives has hardly changed since I bought my first 200Mb scsi drive back in 1992 (1ms track-to-track, 14ms full seek, 10ms random seek), although sequential speed is about 40 times faster.

      The _big_ leaps have been in IO speed(*) + continuing miniaturisation and it's only been flash drives which have kickstarted a new speedup in operational speed.

      (*) Although, doubling a bus speed and then having to add a bunch of waits often means the real speedup is only 25%

      Meantime tape just quietly keeps getting better, but tape densities have at least 2 orders of magnitude to go before coming close to matching current disk platter density. That means there is a lot of room for improvement and there's an easy potential to put 200Tb in a single LTO cartridge - IF head+servo technology can track the improvements in coatings.

  6. Anonymous Coward
    Anonymous Coward

    Photos?

    Quote

    A single photo can be 3-5MB

    What size crappy jpeg are you using?

    Every time I press the shutter on my DSLR I use just over 60Mb of CF/SD card space.

    Perhaps you need a new camera?

    1. foxyshadis

      Re: Photos?

      Good for you, but your camera isn't connected directly to a tape drive, and the intersection of LTO-7 and pro photographers is near-zero. Most businesses only store and back up processed versions of photos from graphic designers and camera phones, not raws. Even Getty doesn't keep raw files, they max out around 10MB (20MP) TIFFs.

      1. Joerg

        Re: Photos?

        "Good for you, but your camera isn't connected directly to a tape drive, and the intersection of LTO-7 and pro photographers is near-zero. Most businesses only store and back up processed versions of photos from graphic designers and camera phones, not raws. Even Getty doesn't keep raw files, they max out around 10MB (20MP) TIFFs."

        There are many photographers with LTO5 and LTO6 tape drives on Mac using LTFS.

    2. phuzz Silver badge
      Meh

      Re: Photos?

      3-5MB? Nah, 1MB is plenty for a jpeg if you're not planning on blowing up to billboard size.

      If I can make it smaller through judicious cropping and colour reducing I will.

  7. Henry Wertz 1 Gold badge

    Movies

    Do I expect random people to be using raw codecs and shooting like 8k 60fps video? Nope. But, I can say at least Hollywood could use a setup like this. The claim is that Avatar takes up 1PB of space in whatever master format they used.

  8. Anonymous Coward
    Anonymous Coward

    Take a step back

    Every time someone talks about tape you get the same old silo'ed views. But the reality is

    Tape will go away when:

    * You want to trust you data to the cloud and allow every hacker who is capable of getting into it to see what you hold dear. If its on the web its hackable.

    * When online storage is so cheap and plentiful that you can have dispersed datacenters that will loose only $1m worth of data if bad thing happen. Based on storage and your ability to replicate large (TB's) amounts of data quickly (minutes)

    * When you believe de-dupe software has no bugs. Please I need to stop laughing.

    * When you forget to deploy a solution to the lifecycle of your data and listen to the sales guys jaw away.

    Tape is meant for the big boys these days. Its expensive, Its not as fast as other storage solution, but lets face it it secure.

    Have you ever read the formatiing in LTO. Its far from trivial. Add to that you would deploy a disc and tape system, with considerable amounts of redundancy designed in and then put your data in a secondary secure location. BTW if you are dumb enough to give said data to a company who employees the cheapest guy they can find to transport your data and then get bitten becuse they have lost it, guess who's to blame.

    In the end of it it come down to one thing how much do you value your data. Then spend accourdingly because if the unthinkable happens you don't want to rely on someone else or on one thing.

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Biting the hand that feeds IT © 1998–2019