Home | Opinion | What’s Wrong with Tape?
What’s Wrong with Tape?

What’s Wrong with Tape?

0 Flares Twitter 0 Facebook 0 Google+ 0 StumbleUpon 0 Buffer 0 LinkedIn 0 Filament.io 0 Flares ×

Last week I briefly attended the Tape Summit at IP Expo.  Unfortunately due to other commitments during the day I was only able to join the round-table.  To be fair this was probably the part of the day I most wanted to attend – hearing everyone’s opinion generates much more interest than listening to canned presentations.

Unfortunately, I didn’t know most of the attendees around the table, however it was clear that there were representatives from the tape industry including LTO consortium members and of course our very own Chris Mellor.  A number of things struck me during the conversation.

The Tape Industry is In Denial – Yes it’s true.  The tape industry can’t understand why end users don’t want to use their products.  There are plenty of arguments around tape being cheaper, easier, having better longevity and so on.  But these don’t resonate with end users who don’t want to use tape when there are alternatives.

Tape Still Has a Place – There’s nothing wrong with tape.  It’s a well-proven, reliable long-term data storage medium that continues to evolve.  Admittedly it’s a case of evolution rather than revolution but that doesn’t matter.  Capacities and throughput continue to rise ahead of the capabilities of individual disk drives.

Usability is all Wrong – Almost all of our tape usage today is through proprietary formats.  For a non-portable medium like disk, proprietary formats don’t matter.  The common interface of NFS, Fibre Channel, CIFS or some other format makes the data accessible.  However tape is inherently a portable media format and so the layout of data on tape itself should be  self describing and independent.

Backup Software Vendors are Holding us Back – Yes, by having proprietary formats used to store data on tape that effectively represents the *same* content i.e. files – tape is losing ground to disk.  Tape requires significant effort to migrate from one media format to another.  Disk systems are much more flexible – think how easy it is to move a backup disk storage device from one place to another compared with moving the contents of a tape to a new one.

So how do we fix things?  The problem is how to correct the usability issue.  We need to be able to move data around between media (e.g. tape to tape & tape to disk) without having to use the backup or archive platform on which the data was created.  Yes, the owning platform needs to know about the change in location if it will continue to own it, but we shouldn’t be restricted to storing data on that software platform if we don’t want it there.  Part of this solution therefore means having decent standards – standards for storing backup objects and archive objects that work cross-vendor and cross-platform.  It’s something that perhaps SNIA should be doing.

We do have one independent format and that’s Linear Tape File System – LTFS.  LTFS seems like a great idea.  It creates a self-defining format for data on tape, embedding the data and the index together.  In this way, data can be read by any platform that simply scans the tape.  Using LTFS wouldn’t stop vendors storing the data within files in their own format, but it would at least give us a fighting chance at consistency.

So, end users, tape manufacturers, lobby the backup software manufacturers – some of which are the *same* company (e.g. IBM) and start creating consistent open formats for tape storage and users might start to think that tape was worthwhile and so there might be longevity in your products after all.

About Chris M Evans

  • chriscowleysound

    The proprietory nature of tape is always something that niggles me. My old place had tapes going back to the early 90’s, but the did they have the ability to read them? Of course not, yet I could take the hard disk from my dad’s old 386 and plug it into a modern PC – most likely without the need for an expansion card! This maybe for software reasons, or hardware.

    And another thing! (call me Mr Clarkson) What was wrong with the tar format? It was missing things (such as ACLs), but it was BSD licensed, so any company could have just added an extension to enable what they needed. Ok, maybe I would have lost the ACLs (I’ll continue with that example) if restored with the wrong version of tar, but my data would not have been trapped.

    • http://thestoragearchitect.com/ Chris M Evans

      tar would work (in fact doesn’t Netbackup still use tar?). As you say, ACLs are a potential issue on restoring data, but you are right that could easily be overcome. I’ll say again, I think the tape industry needs to take the lead and develop more standards to make their products more consumable.


    • DeepStorage

      No you couldn’t take that ST-506 disk from your 386 and put it in a modern PC with only a SATA interface. You’d need an old DLT7000 drive to read a DLT7000 tape but at least that would be SCSI and you can get a PCIe SCSI HBA.

      No one ever made PCI or PCIe ST-506 controllers so you’ll need an old computer with an ISA bus slot for the disk controller and then need to enter the drive geometry (number of heads, tracks, sectors per track and the interleave used when the data was written) into the old version of DOS (or Windows NT 4.0 or earlier) to read the disk.

      LTFS is a big step up from TAR but the backup vendors have reasons, granted many about customer control, for proprietary formats.

      • http://thestoragearchitect.com/ Chris M Evans

        Howard, it’s easy to take the extremes and obviously an original ST-506 isn’t going to work today. However ATA has been around for many years and that and SATA have been ubiquitous as consumer connected devices. So there are many drives out there that can easily be read.

        Consider LTO; Gen-1 for read/write, Gen-2 for read-only and Gen-3 downwards you’ve no chance, so reading even tapes from 5 years ago could be an issue unless you retain multiple drives.

        Ultimately as I mentioned, disk formats are easier to work with as they are standard. Tape tends to have proprietary formats applied. As you mention, it’s about control – maybe there’s an opportunity to develop an Open Source backup tool/format.

        • DeepStorage

          You’re conflating a couple of things Chris. Because today’s disks (basically everything since that ST-506 including PATA, SATA, SAS and SCSI) combines media, drive and controller in a single package the physical side of the format is settled in the one package. On the other hand the file system and RAID layout of the disk are different for each system you put that drive into.

          If you take a disk drive off the shelf you have to guess what file system the disk was formatted in. If you take an EXT2 disk and stick it in a PC running Windows it’s just as greek as if you try to read a Networker tape with Backup Exec. (BE can read old Arcserve tapes just as a Mac can read NTFS)

          Think about the difficulties of having a RAID set of 3.5″ disks from a Celerra and reading them on a NetApp?

          Snice the world standardized on LTO life has gotten a lot simpler. HP and IBM still make LTO-3 drives so every LTO tape ever written can be read with NEW drives.

          If we have a need for removable storage, and services like Amazon’s Glacier make that questionable for many people, we’re going to have to track where the media came from and where it can go back to.

  • jason

    Crosswinds came to us (media company) pitching an LTFS backend, with a CIFS/NFS frontend, with a smallish (2-3tb) on disk cache. It’s a great system (in theory), but it’s priced by the TB of space used, which is STUPID. It costs more to buy the capacity for the tape, than the tape cartridge itself (by a large margin).

    That would be a great tool in the arsenal… It surprises me that there isn’t an open source way to do the same thing already…

    • http://thestoragearchitect.com/ Chris M Evans

      Jason, good point about the open source idea. I guess to many, tape just isn’t sexy enough to invest in developing stuff for free. This is perhaps where the tape industry could step in and do something. For a little work, they could get tape integrated further into many environments by default.


  • Pingback: Internets of Interest for 1st November 2012 — EtherealMind()

0 Flares Twitter 0 Facebook 0 Google+ 0 StumbleUpon 0 Buffer 0 LinkedIn 0 Filament.io 0 Flares ×