Tape still going strong

As a person who has spent a good portion of the last couple of years talking about disk backups, one might think that I’ve given up on tape.  Nothing could be further from the truth. In fact, I’m encouraged by my friends over at Spectralogic that just released the industry’s largest tape library.

I’ve been thinking about this post ever since I saw the press release on the T-Finity tape library over at Spectralogic.  It’s got over 30,000 slots and 480 tape drives!  But it was a comment on my last blog post that actually got me typing.   My friend Jay over at SEPATON said in his blog post about the T-Finity that disk is now the primary target for backups and that tape is used for long-term archive.  I agree with Jay that this is how I would design a system today, but that still leaves room for a lot of tape.  Let’s just say that if there are customers that need an 8-node SEPATON system, I think there are customers that need a 30,000-slot tape library.

(I remind anyone who’s curious that I am not being paid by anyone to write this blog.  This is just my opinion.)

First let me say that even with the advent of deduplication, tape is still the cheapest medium on which to store backups and archives for the long term.

Before looking at long-term storage, let’s look at onsite backups.  When we look at nearline storage systems (e.g. deduped disk system vs fully-loaded tape library), deduped disk is close to the price of tape, but tape is still cheaper.  Maybe in a competiive pricing war, disk might win out because the deduped disk guys have more margin to play with than the tape guys.  But list to list, tape wins in nearline storage.

But where tape really shines is in long-term retention.  If you want to keep 7+ years of archives on tape, and I don’t plan on reading them much, you can’t beat tape.  On a shelf, an LTO-4 cartridge costs me $0.03/GB ($35/tape / 800 GB * 1.5:1 compression)! Even the cheapest disk drive I can buy at Fry’s (2 TB disk for $179) is three times that. And since that disk drive is not made to sit on a shelf powered off for years at a time, that disk will need to be in a disk system (typically at least quadrupling the price) and powered on for 7+ years!  Imagine the cost of that power vs the cost of that tape sitting on a shelf.

As to tape’s reliability, I’ll take long-term data that was properly stored on a powered off, disconnected tape over data on a powered-on, easily-hacked disk any day of the week. In sixteen years of doing backups all over the place, I’ve had far more disk drives fail on me than tapes.

Where tape has become a pain is as an initial target for backups.  It’s nearly impossible to make a 180 MB/s tape drive happy with backups coming directly from their source to tape. You can do it if you stage them and serialize them on a high-end disk system, though.  That’s why so many people have gone to disk-to-disk-to-tape backups.  And many have also gone to storing all of their onsite backups on disk using deduplication.  But the bulk of people are still making tape copies for offsite storage.

If you do plan on storing data on tape long term, you’re going to need to do a few things.  (This is in addition to the obvious of it being stored properly in a humidity-controlled environment.)  The first thing is that you’re supposed to re-tension it every year.  That is, put it in a drive, fast-forward all the way to the end, then rewind, then put it back on the shelf.

The second thing you’ll need to deal with is changing formats. Disk does make this really easy (assuming it’s not a VTL or a CAS device) by allowing you to do a simply copy from the old filesystem to the new filesystem.  On tape, you’ll need to occasionally (depending on how often you swap drive types), copy all the old media to newer media.  My personal preference is no less often than 5 years.  (BTW, I’ve never met anyone that does this systematically.)

Finally, anything that’s worth having on a tape for seven years is worth having on two tapes for seven years.  Please have two copies of everything stored in two different locations.  (Yes, that doubles the cost, but it’s still cheaper than disk.)

What about Magneto-Optical, you say?  I’m glad you asked. There’s no doubt in my mind that MO is a much more stable platform for long-term storage, but it sure isn’t a cheap one. Are you sitting down?  A single 5.2 GB optical disk costs about $35! That’s about $7/GB, over 200 times the cost of tape.  Holy cow!  It makes a great long-term storage mechanism (and doesn’t need the retensioning/recopying stuff we just talked about), but man that’s expensive.

Tape’s not going anywhere any time soon.  And I congratulate Nathan and crew at Spectralogic for leapfrogging the big guys with the biggest, baddest, tape library on the planet — not to mention the densest. It’s amazing how these guys figure out how to fit so much in such a small space.

Written by W. Curtis Preston (@wcpreston), four-time O'Reilly author, and host of The Backup Wrap-up podcast. I am now the Technology Evangelist at Sullivan Strickler, which helps companies manage their legacy data

3 comments
  • Just a small comments.

    From what I know Sun SL8500 is the largest tape library. It can contains 70,000 cells with up to 448 drives when 7x SL8500 are grouped together. Plans are to extend to 10x SL8500 soon. Sun own that market since years.

    GREAT comments on your tape analysis. Yes tape beat disk once again…wait next year when they will double or more the capacity with new media formula. Some will storage up to 2 to 4 times more than what they actually do today.

    Yes tape drive speed is a problem. Media handling and management is another one too. That is the problem with tape. Other than that life with tape is much better than disk. Having to deal with SATA failure and huge potential data lost ( including dedup) you really start to like tape…the golden copy who saved thousand of admin per year…

  • (Their fault, by the way. I’m more than happy to take a briefing.)

    From what I remember (and I prepare to stand corrected), the 8500 is the largest tape library SYSTEM. But….

    1. It’s actually not A tape library. It’s SEVEN tape libraries connected with pass-thru slots. Anyone that knows what they are knows that is a far cry from what Spectralogic has come out with.

    2. IIRC, you need ACSLS to present the multiple libraries (because they ARE multiple libraries) as a single virtual library to a single host. Yes, I know ACSLS does more than that, but my understand it is required to present these multiple libraries as one library.

    Please correct me if I’m wrong.

  • [quote name=W. Curtis Preston](Their fault, by the way. I’m more than happy to take a briefing.)

    From what I remember (and I prepare to stand corrected), the 8500 is the largest tape library SYSTEM. But….

    1. It’s actually not A tape library. It’s SEVEN tape libraries connected with pass-thru slots. Anyone that knows what they are knows that is a far cry from what Spectralogic has come out with.

    2. IIRC, you need ACSLS to present the multiple libraries (because they ARE multiple libraries) as a single virtual library to a single host. Yes,ccna (http://www.myccnatraining.com) I know ACSLS does more than that, but my understand it is required to present these multiple libraries as one library.

    Please correct me if I’m wrong.[/quote]I think same as you thinking.I think there should be more such tape libraries.