Tape vendors thrust LTFS tool at punters: Go on, you know you want our tape
Reclaiming the tape software interface
IP Expo: Tape Summit Two years after its inception, the Linear Tape File System (LTFS) is the tape hardware industry's best hope to claim a piece of the tape software interface and repair the damage done by decades of proprietary and hard-to-use backup software.
We've all heard that "tape is dying" from the various disk array vendors, while others chime in: "you need disk for restore speed" and "deduped disk is as cost-effective as tape". We've all drunk some of the D2D (disk-to-disk) backup Koolaid, but that doesn't make it true.
We'll get to that in a minute but, first, let's examine LTFS. It was invented by IBM engineers, announced in 2011 and provides metadata about a tape cartridge's contents on one of the tracks. This metadata describes the files on the tape and the folders they logically inhabit. A host computer can use LTFS to have a file:folder GUI interface to a tape drive and a cartridge loaded in it.
Users can find out what's on the tape, drag-and-drop files from disks on their system to the tape and drag and drop files in the tape to their local disks. Media files on the tape can be played on a host, and this can be done without using tape backup software so loathed by generations of users who have myriad horror stories about their experience using tape backup software on their PC or workstation. In this area tape so did suck and millions of users became fervent adopters of external disks for backup; no backup SW impeding them, and fast and simple file:folder interface operation. What's not to like?
Proprietary backup software
The thing is, from the LTFS point of view - all of this section was written assuming an LTFS point of view - tape backup software is a major hold-out bastion of proprietary, lock-you-in-for-ever software. Operating systems own the basic file:folder disk interface; tape backup software vendors each own their proprietary, hard-to-use, complex bit of the tape interface, not the operating systems and certainly not the tape hardware vendors.
The backup software vendors had no interest, none whatsoever, in creating a standard backup-and-restore interface to tape. For them such a rising tide would not float their boats but leave them stranded and open to the bitterly bracing winds of open competition.
This did not matter to the tape hardware vendors, much, when there was no alternative to tape for backup and archiving. But ten or more years ago deduplication was invented and suddenly backup software found itself writing to deduped disk because it was quick enough for writing and vastly faster for restoring files, a random operation, than poor old sucky sequential tape.
Startups like Avamar, Data Domain, Diligent, Exagrid and Sepaton - "No tapes" spelled backwards - assaulted the tape backup market, vowing to kill tape backup, and found they were knocking at a wide-open door; customers virtually pulled them through. The startups prospered; the mainstream disk array and server vendors saw the growth potential and bought their way in. The disk-to-disk (D2D) backup business boomed, with virtual tape libraries and NAS interfaces added to the mix. It became a multi-billion dollar industry and every dollar spent on deduped disk arrays doing backup was a dollar that could have been spent on tape backup.
To a large extent the backup software vendors caused the problem for the tape vendors. Not wholly, because there was no standard tape format; you couldn't read a VXA tape in a Mammoth drive, to state the bleedin' obvious. There was no standard tape interface and no alternative to using backup software to restore files. Speaking from personal experience with a DAT tape and a nameless software vendor's backup software sucked, it really did suck.
A tangle of tapes
Birth of LTO consortium
But some tape hardware vendors learned and IBM, HP and Seagate – odd that as Seagate is a disk supplier, but it had a Certance backup business back then - set up the Linear Tape Open (LTO) consortium and standardised the Windows/Unix/Linux tape hardware market.
LTO tapes could be accessed on any of the LTO vendors' tape drives. Competing formats, like DLT, SDLT, A-AIT, VXA and others slowly collapsed and died. Now HP, IBM and Quantum, which bought Certance, the LTO trio, rule the tape hardware roost with just Oracle and IBM left with proprietary tape formats for mainframe customers.
But the tape backup software vendors did not appear to learn from this, and this continuation of proprietary and complex tape software and slow tape hardware helped widen the opening for the deduping to disk arrays and VTL vendors who made disk arrays look like tape drives and enabled tape backup SW to write to disk much faster than to tape. And when mainstream server and storage vendors bought deduping array start-ups, the rot began. "Tape sucks" marketing messages flooded media channels and deduping arrays poured through the breach in the tape HW/SW dyke.
Tape market chart with projected revenues tailing off to death. Wrong.
Conceptually the war is won, it's commonly thought now that tape is not for backup, that this is disk vendors' fief, with tape relegated to archiving cold data, the place data goes to die. Well, it seems to be that way, but actually it's not.
Tape lives on
Most small and medium enterprises (SMEs) still use tape for backup. IBM's Paul Scheuer, the programme director for tape brand management in IBM's storage systems division, said a Gartner study in 2011 found only one in four backup users have moved entirely to disk backup away from tape – 27 per cent to be precise.
Yes, a huge amount of deduping disk gear has been sold but the bedrock of SME backup is tape. The dedupers see a lot of opportunity here; however they are fighting an uphill battle. Tape is cheap and tapes sitting in a car boot or on a shelf consume no power. Disks spin, and however you spin that it means electricity is being used.
An Alpha TV study in Greece, looking at video data storage, found LTO-5 tape storage cost $0.04/GB whereas one disk alternative cost $0.10 and another, XDCAM disc, was $1.35/GB, a 37X increment. Now we have LTO-6, which holds more than LTO-5, so the cost is lower again. A Clipper Group report found:
Disk is more than fifteen times more expensive than tape, based upon vendor-supplied list pricing, and uses 238 times more energy (costing more than the all costs for tape) for an archiving application of large binary files with a 45 per cent annual growth rate, all over a 12-year period.
That's for archival purposes and not backup but you get the picture. And there are other hammers we can use to hit this particular nail. Scheuer said an NSIC study showed tape storage used 96 per cent less energy over five years than disk. He says that tape's hard error rate is less than disk: "It's why disk is RAIDed."
Tape use is widespread, with 73 per cent of businesses and organisations doing backup using it. It's cheap, getting cheaper, and it's even green. That's an opening for LTFS; remove the dead proprietary hand of the backup software companies and provide an instant, familiar, easy-to-use alternate access method. As ESG's Mark Peters said: "Thanks to file systems such as LTFS, tapes can look and act like enormous USB drives."
IBM 35TB tape test rig
Tape areal density increase isn't slowing and isn't facing expensive technology upgrades like disk as perpendicular magnetic recording technology (PMR) meets its limits. That's because a 1TB tape has a 1.2Gbit/in2 areal density while a modern hard disk as at 635Gbits/in2. The tape bit occupies a physical space of 8,000 x 65nm whereas the disk bit takes up 74 x 13.5nm. That means enormous scope for areal density increases.
IBM demonstrated a 35TB tape in 2010 and is working on a 125TB tape technology demonstration.
IBM has made the LTFS access method openly available. Scheur's Tape Summit presentation stated: "Many tape storage companies offer "freeware" versions for Windows, Mac OS, Linux and other operating systems."
Existing tape use cases are backup and restore, disaster recovery, offline archiving and long-term data preservation, and traditional big data storage. LTFS' occurrence is driving new use cases such as:
- Active archiving with a blend of disk and online tape
- New big data storage for the media and entertainment, digital video surveillance and analytics markets
- Cloud storage - think Amazon Glacier
IDC predicts the amount of data kept on tape is growing and nit declining. It expects an annual tape data storage CAGR of 45 per cent from 2010 through 2015.
The hot topics in storage are widely thought to be flash, the cloud, and virtualisation, with disk becoming a little stodgy and tape … well, yes, tape, isn't tape dead on its feet? No, it is not. Data lives in a spectrum. At the fast access end we have flash storage taking over from disk. At the cold data, slow access need end we have seen disk mount a raid on tape, but that raid is running out of steam, and tape's virtues are being recognised afresh.
Tape as a long-term data storage medium is far more cost-effective than disk, even deduped disk. And tape with LTFS presents a gigantic backup and archive resource for individuals and businesses that need easy access to huge datasets that are too large for a disk. El Reg storage desk has a new mantra: "Tape rocks". ®