Can you hear bells tolling? We've recently written about two cusomers using tape for video storage because nothing else comes close to its cost per GB stored.
SpectraLogic has just added two frames to its T950 library and raised capacity to 10,050 slots and 16.08PB - if you use SAIT tapes. Tape automation vendors are banging their drums, the tape market looks healthy and yet, if you listen hard, you might hear bells ringing.
Overland Storage increased has just increased its REO virtual tape device speed five-fold through a more powerful CPU and flash memory cache. Try doing that with a tape drive. The tape backup window problem is insoluble - with tape. It's all too soluble with disk though and every backup software vendor and tape automation vendor knows it. There are examples all around. SpectraLogic includes a virtual tape library with its T950, the removable RXT system. Tandberg has a removable disk backup product using ProStor removable disks. Yosemite Software alters its backup software so it will backup to disk. Microsoft has its DPM disk-to-disk backup software. Talk privately to tape product vendors and they will admit that the SME Market is a lost cause.
Every tape vendor in it is pretty much resigned to losing the SME market over the next few years. In the mid-range space disk is appearing as a front end backup cache for longer-term offsite tape backup. But it could do more. De-duplication holds the promise of radically increasing the amount of backup data that can be stored on disk, by factors of twenty or thirty when repetitive backups are run. A 500GB tape reel can hold a gigabyte of compressed data. A 500GB disk could hold, the de-dupe boosters say, effectively 10GB of backup data, perhaps even more.
What does that do for the cost/GB number for disk?
Now add removable disk to the equation, and we could envisage an automation device that uses robotics to move disks between slots and drives, just like a tape library. That's technology fantasy at the moment but it's not that far away; Imation has removable drive technology that it is pushing to potential OEMs. With this, tape drives in a library could be replaced by disk drives emulating tape drives and backup and restore speeds increase commensurately.
Of course, with MAID, the technology that spins down or idles inactive disks, you don't actually need robotics. Idle drives just sit there waiting for electronic wake-up calls. Copan pioneered it. Nexan followed suit. Others are now looking at it. A MAID array could hold more data than a tape library in the same physical space. A de-duped MAID array could hold vastly more and its cost per GB stored could be, might be, less than tape library. That would be a huge barrier breached.
Customers with mid-range and enterprise tape libraries are very conservative and represent a profitable installed base for library suppliers. They don't switch formats from 9940 to LTO or to SAIT. They don't switch library suppliers. But they might switch to disk if its speed advantage came without an increase in the cost of storing each gigabyte. They might be even more willing to consider MAID disk once they consider the power and cooling implications. Idle disks don't consume power and don't need cooling. Have your backup window solution cake and eat your low storage cost and green storage cake at the same time.
We can paint a picture showing how slowly but steadily the outstanding advantages of tape technology are being whittled away. If you are a tape automation vendor ask not for whom the bell tolls. It tolls for thee. Unless you have your head in the sand of course. Then you can't hear it - but you could still get your butt kicked.