Do We Need 35TB Tapes?

Last week IBM and Fujifilm announced that they've developed a new tape technology that could hold a whopping 35 TB on a DLT/LTO size cartridge before your data supposedly compresses 2:1 to put 70TB in the palm of your hand. While I'm impressed with the technical feat, I'm wondering what we would use a 70TB tape for.

The first problem will be just keeping the beast fed. IBM and Fujifilm didn't say anything about performance, but for it to be a useful backup medium, a 35TB tape drive would have to be significantly faster than the 180MB/s that an LTO-5 drive promises. At that rate, it would take 55 hours to fill a tape, which is a lot longer than most backup windows.

In fact, to get the full tape recording time down to a reasonable eight hours, the new tape drive would have to suck in data faster than the top of the line Data Domain DD800. Even a two-drive library would need 40Gbps Ethernet connections just to avoid shoe shining. In fact, I wouldn't be surprised if the SpectraLogics of the world built a new generation of libraries complete with SSD caches big enough to write a full track of data.

With the right drive technology, really high density tapes make a lot of sense for long term archival storage. If MGM hadn't been foolish enough to sell their library to Ted Turner years ago, they could keep all that footage in a single PowderHorn-sized library and send additional copies to undisclosed locations for safe keeping.

That said, the materials science technology is impressive. By using ultra-fine, perpendicularly-oriented, barium-ferrite magnetic medium, they're not only getting staggering data density but also avoiding expensive metal sputtering or evaporation coating methods. Theoretically, the new tapes should also be more stable than today's metal particle technologies, since BaFe, despite the abbreviation leaving out the O for Oxygen, is an already oxidized ceramic that won't need the passivation layer that metal tapes need to protect them from oxidizing.The road to market acceptance is rocky however. One possibility is this technology goes down the road of holographic storage, and the periodic demonstrations of high density optical drives remains a science experiment forever. The other dead-end would be for IBM to try to recoup the development costs over too small a pool of sales, pricing the technology out of reach of most organizations.

Unlike those in the "tape is dead" camp, I believe in tape on the shelf as an archival medium. Even with today's technology, you can put more data in less floor space with tape than with alternative media, and of course, tape on the shelf doesn't draw any power or need special cooling. If we're lucky, by 2020 we'll be able to buy a 35TB tape for $400 or so, and I can put the entire Dr. Who series on a single tape for posterity.