back to article Tape vendors feel the cold, clammy hand of AWS on their shoulders. Behind them grins the Glacier Deep Archive

A bunker-busting bomb just exploded in the tape business. Amazon today teased during its re:Invent conference in Las Vegas a deep cloud archive storage system that will be a quarter of the price of its existing Glacier system. Glacier is Amazon’s original cloud archival storage offering, and costs $0.004/GB/month ($4/TB/month …

  1. Mayday

    Retrieval time

    The most important thing with a managed backup solution isnt it?

    Of course bean counters like the cost, but I want to know can I get my data back (does your DR system actually achieve this?), and how long will it take?

    Glacier does what it says on the pamphlet of course, and Im a fan of how you can move stuff there from standard S3 automatically after a certain time.

    1. Paul Crawford Silver badge

      Re: Retrieval time

      And said time also depends on your bandwidth and the volume of data to be restored. Unless you are just re-populating an AWS instance, of course.

    2. WhackWakie

      Re: Retrieval time

      Maybe, soon company will keep only latest backup copies locally on QLC flash, and also push the backup into the cloud. If the local DC burnt down, I doubt local backup will be urgently needed.

    3. JohnFen

      Re: Retrieval time

      "The most important thing with a managed backup solution isnt it?"

      The most important thing is the availability and integrity of the backups. Since retrieving the backups is (or should be, anyway) an extremely rare event, the amount of time it takes isn't as important (although faster is generally better in all things, of course).

      1. David Harper 1

        Re: Retrieval time

        "the amount of time it takes isn't as important"

        I'm guessing you've never had a pointy-haired boss screaming at you when someone accidentally dropped the company's mission-critical database, and you're the DBA who has to get the whole thing up and running again from the backup tape. Retrieval time seems pretty fscking important then :-)

        1. JohnFen

          Re: Retrieval time

          "Retrieval time seems pretty fscking important then"

          I didn't say it wasn't important, just that it's not the most important thing. The pointy-haired boss may scream that you aren't restoring the system as fast as he'd like, but he's more likely to fire you if the backups are screwed up.

    4. Korev Silver badge

      Re: Retrieval time

      Expedited retrievals cost $0.03 per GB and $0.01 per request.

      Standard retrievals cost $0.01 per GB and $0.05 per 1,000 requests.

      Bulk retrievals cost $0.0025 per GB and $0.025 per 1,000 requests.

      Looks like ~256USD to bring back 100TB; however what counts as a "request"? Is each file one request?

      1. Anonymous Coward
        Anonymous Coward

        Re: Retrieval time

        "Is each file one request?"


      2. Spazturtle Silver badge

        Re: Retrieval time

        "Looks like ~256USD to bring back 100TB;"

        You have made the same mistake quite a few people have made and been stung by, use the price calculator:

        Retrieving 1TB that was stored for 2 years will costs $1887, retrieving 100TB that was stored for 2 years would cost ~$200k.

        The calculator doesn't include request costs so the actual amount will be higher.

    5. Anonymous Coward
      Anonymous Coward

      Re: Retrieval time

      "The most important thing with a managed backup solution isnt it?"

      Glacier is the equivalent of *off-site* tape backup. That means it should be your last-resort restore point (i.e. rare occurance), and generally isn't quick (due to the need to actually get the tapes from a remote location, etc)

      1. disk iops

        Re: Retrieval time

        > (due to the need to actually get the tapes from a remote location, etc)

        I hope you're not suggesting Glacier is tape-based. There isn't a tape drive to be found in any of their AWS data centers. Tape drives are unreliable as hell. And the data-cartridges are barely any better.

        The 'nines' thing is ridiculous, I agree. I very much doubt they are calculating those availability numbers in the event a few of the regional datacenters get nuked or otherwise suffer EMP or other massively disruptive event. Granted who cares at that point if your data still exists...

        Typical S3 requires half of the erasure-coded blocks to be available. Legacy Glacier used the same N:M ratio because they wouldn't have to reconstitute the data in order to store it. The new service may well change the N:M ratio but I rather doubt it. Instead of 3 datacenters they may have bumped it to 5 and/or transparently copied the a minimum 'N' to an alternate region. Most likely the cost savings are achieved by using ever larger disks (eg. 12+TB vs 4TB) and ever larger disk enclosures (used to be 96 per tanker) which is probably closer to 240 and maybe even higher.

        The tiers of service time is just a job-sorting/priority mechanism. If you buy 'fast' restore, your job gets put at the top of the heap to get scheduled. If you have medium and your window is closing, it job gets bumped up so it can complete within SLA.

  2. hellwig

    Cloud based archiving?

    I already shipped my tapes to Iron Mountain so they could upload my data to Google. Now I'm gonna need those tapes back. Great.

  3. Pascal Monett Silver badge

    "13 nines durability"

    Did you hear that rushing water sound ?

    That's the sound of millions of Microsoft customers salivating at the prospect of a cloud that actually helps them, instead of keeping them from working.

    But hey, Nadella, stay the course ! Nothing like being a living promotion for Open Source tools, right ?

    1. Mayday

      Re: "13 nines durability"

      These figures (ie a swag of nines) I dont quite get how they quantify. Different amounts for availability and durability but still more 9s than I have fingers to count them on.

      Wonder what Alexa Skill they used to figure it out?

      1. JohnFen

        Re: "13 nines durability"

        "13 nines" is simply meaningless marketing wank.

  4. Tomato42


    That's actually cheaper than Backblaze.

    Depends on retrieval cost and delay Amazon may finally have something competitive

    1. Spazturtle Silver badge

      Re: Impressive

      The retrieval and transfer costs are higher then for Glacier, so look at the Glacier prices and realize that this is much more expensive then backblaze:

      1. Tomato42

        Re: Impressive

        doesn't that calculator use the current Glacier prices, not the one from the article?

  5. trevorde Silver badge

    There is no such thing as the glacier...

    ... it's just someone else's tape archive

  6. Ima Ballsy

    Hot Damn ....

    A cheap place to stash all my P0rn now !!!!

  7. Anonymous Coward
    Anonymous Coward

    I would love to see their service engineering docs

    "13 nines durability; 99.99999999999 per cent, compared to Glacier's 11 nines"

    If we ignore for a minute the question of how they got to the original 11 nines for Glacier...

    What I really would like to know is did they really sit down and say "ok guys, how do we add another two nines ?"

    True 11 nines would already involve a substantial degree of over-engineering and redundancy. So I am genuinely curious as to what exactly would need to be done to add the extra two nines ?

    1. doublelayer Silver badge

      Re: I would love to see their service engineering docs

      Engineer: Well, we capped it at eleven nines before for an event that has a likelihood of occurring once in a hundred billion years, like the sun going red giant on us. That would mean that our geographically redundant system wouldn't help much.

      Marketer: What would you need to add some nines to that number?

      Engineer: Well, we already added one last time on the assumption that we'd have travel outside the solar system by that point, so a really cataclysmic thing would only happen one in ten times a star destroys all its planets. And because you wouldn't stop badgering us. So we can't really--

      Marketer: [interrupting] We need something new to distinguish our new product from our old product.

      Engineer: Well, it costs less. How about that?

      Marketer: But it's also better, right?

      Engineer: We already have tapes with redundant copies stored in libraries on six continents, in a total of eighty nine datacenters, in order to let me sleep soundly with all the nines we have now. So we did open a few more datacenters to store the tapes, but not really enough to make the number any longer. Also adding more nines would be pretty pointless.

      Marketer: I give up. We'll think it over and see what can be done.

      Accountant: Just slap two more nines on the number; it's meaningless anyway. We haven't listened to the engineers for years. They still think we have eighty something datacenters all over the world.

      Marketer: What do we actually have?

      Accountant: A warehouse in South Dakota.

      Marketer: So our real level of reliability is?

      Accountant: It's a pretty good warehouse. Maybe 99.9% or so.

      Marketer: What happens when people find out?

      Accountant: I'm still paying that engineer, the one who's in charge of and thus responsible for our entire glacier system, aren't I? Don't worry, I've covered every contingency. Our employment and liability is 99.999999999999% secure.

      Marketer: I should probably quit before something happens, right?

  8. G Olson

    Blah, blah, blah...these cost comparisons are always so shallow. I just created a cost comparison of LTO7 tape to AWS for the seven year life span of the tape array. Even with the new Deep Ice, local tape is still cheaper. And the bonus which is never included -- those tape drives will last for 15 years. I can move the primary tape array into a secondary role for another 7 years and be more cost effective. And I won't have to worry about API Of The Month Club changing my access method.

    1. Anonymous Coward
      Anonymous Coward

      "I just created a cost comparison of LTO7 tape to AWS for the seven year life span of the tape array. Even with the new Deep Ice, local tape is still cheaper. "

      I don't know about enterprise-quantities of data, but for home users with "a lot" of important data (i.e. photo-hoarders, aka hobbiest photographers), it'd take 11 years for 1TB of Glacier to cost the same as a 2-bay NAS with mirrored 1TB HDDs (not factoring in replacing the HDDs), or around 5 years for 2TB. It works out significantly cheaper than tape for us (given the cost of tape drives). ;)

      1. Anonymous Coward
        Anonymous Coward

        "11 years for 1TB of Glacier to cost the same as a 2-bay NAS with mirrored 1TB HDDs "

        I wish I could give you a million thumbs down.

        At least make a sensible attempt at a comparison !

        Your "a 2-bay NAS" is equal to ONE copy. I don't care what RAID you're using, RAID is ONE logical copy.

        If you are going to make a sensible comparison to AWS, then you should **AT LEAST** start with the basic old-school rule of thumb, namely:

        "Three independent copies. Preferably *at least* one of which is always located at a geographically dispersed location to the other two."

        So, how about doing your numbers again with **three** "2-bay NAS with mirrored 1TB HDDs" ? Of course not forgetting that AWS automatically gives you the geographic replication, so you'd better add some nice high bandwidth capacity between your sites.

        I'm not saying Amazon is always the cheapest or the best. But for hobbyists who think RAID gives you extra integrity, I would say Amazon is probably a safer bet !

      2. doublelayer Silver badge

        That's true, but glacier provides a lot of stuff that those hard drives don't. Among other things, I don't trust a two-bay RAID system. That looks like a really easy way to lose both disks, because one will fail and the other one would probably fail too when you use it to mirror again. It also doesn't protect against something that takes out your NAS, such as flooding, fire, lightning strike on the power or data line, theft, or dropping it. I'm not always advocating for cloud backup here; I tend to put a lot of backup data on hard drives which I store off site, but the two are not comparable.

        1. JohnFen

          "I tend to put a lot of backup data on hard drives which I store off site"

          Hard drives are a terrible backup medium.

          1. doublelayer Silver badge

            Hard drives have their downsides as backup media, but they are the most available and functional media for personal use. I do not have enough data to justify purchasing a tape drive, because I have not seen one that is sensibly priced for home use. So what other choice is there? I could use optical media, but that degrades as well and even the largest capacity conventionally available, blueray, isn't very big when compared to disk. I could use exclusively SSD storage, but that would be much more expensive, and can also fail. So my backup strategy uses disks. They are independent such that I can sustain failures in some of them without losing anything, and I don't have a better option. If there is something I haven't thought of that would be feasible, I'd like to hear about it. For now, I have a system which is pretty reliable.

  9. JohnFen

    Still too expensive

    "Amazon today teased during its re:Invent conference in Las Vegas a deep cloud archive storage system that will be a quarter of the price of its existing Glacier system."

    That's still too expensive. But, I confess, I would consider it too expensive even if Amazon weren't charging money for it at all. Give me the tape drives instead.

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

Other stories you might like