Can tape still win in a cloud world?

There’s a lot to love about tape.

For one, it’s durable: archival tape is rated to last up to 30 years. Tape also doesn’t consume power while it sits on a shelf.

And tape is cheap. A $70-75 list price for LTO-6 works out at about three cents a gigabyte. Consumer-grade 2TB SATA drives are only now starting to approach that sort of price, while new versions of LTO will improve tape pricing even further.

But there are downsides. A lot of them.

Tape is sequential, so it’s not suitable for random access to data. Tape also requires special hardware, including custom robotics to move tapes in and out of drives, and obscure software to manage it all.

The ubiquity of spinning media means using disk drives is something anyone can easily do, and there are standard communications protocols to talk to them, from USB to fibre channel. Device drivers for disk are abundant, while tape drivers are specialised beasts that few people know how to program.

The end of tape

The operational issues with managing tape are such a major headache that many organisations are choosing to avoid tape completely and look to cloud services for archiving data.

Or at least, they assume they are avoiding it.

Enter Amazon Web Services and its Glacier archival storage service.

Recently introduced to the Australia region, Glacier provides a low-cost, high-latency storage service available over the internet or private leased lines.

Storing data in Glacier costs a ridiculous one cent per gigabyte per month (accessing it costs an arm and a leg). Crucially, access to the data takes "three to five hours", according to Amazon's FAQ.

For these reasons, most storage engineers are convinced Glacier is little more than one big tape library, with access to it charged as-a-service.

We've asked Amazon point-blank on a number of occasions whether that's the case and every time it has refused to "discuss implementation details".

Tape is dead? Doesn't sound like it.

Abstracting archive

In any case, the AWS strategy is to do all the “undifferentiated heavy lifting” for customers, said Simon Elisha, principal architect at Amazon Web Services.

If AWS gets its way, customers won't ever again talk about the 'information management lifecyle'. Rather than try to classify data and storing it according to the most suitable medium, "they can keep everything, and use it only when they need it,” he said.

AWS takes care of the operational complexities of how the data is stored, freeing customers to concentrate on how they use it.

Take Australian company NearMap, for example. NearMap uses the AWS Simple Storage Service for primary storage of mapping data, and uses Glacier’s automatic life-cycle policies to move data off to Glacier for cheaper storage when appropriate.

Earlier today, AWS announced the AWS Storage Gateway could now be configured as a virtual tape library, with Glacier sitting in the back end as a 'vrtual tape shelf', helping to ease the transition of data from your old tape library to Amazon's.

Interestingly, the tape vendors themselves view new opportunities in a cloud world.

High-end tape system vendor Spectra Logic has written an API for its BlackPearl appliance so developers writing to Amazon S3 storage can write to their own physical tape library for the same application. Dubbed 'Deep Storage', or DS3, the API provides extensions to the S3 API for dealing with large sequential reads and writes.

The BlackPearl appliance thus handles all the communications with tape drives and silos, providing a simple interface to the data in the object form familiar to Amazon S3 developers. Instead of requiring separate backup and archival software to manage the tape infrastructure, application developers can access data directly.

Molly Rector, chief marketing officer at Spectra Logic, told iTnews “per gigabyte, tape has always been cheaper, but the software to get at the data on tape has been more expensive. Once you get to 500 terabytes or a petabyte, the cost of the software to move data to and from tape is more than the cost of the storage itself."

With that compexity abstracted, the game changes. One of the first named customers for Spectra is none other than Yahoo!.

“The ten largest web-scale, cloud-type companies have, by-and-large, [previously] written their own tape integration layer,” Rector said.

“You’ll see [Amazon] Glacier type offerings coming out from a lot more of the cloud service providers,” she predicted.

Taped shut

Most of those cloud service providers will, like Amazon, prefer that users (a) don’t know the technical underpinnings of a service and (b) don't care.

And in many respects, that's probably a good thing.

It's just ironic that the marketing folks at Amazon want you to replace your "cantankerous mechanical devices with scads of moving parts" with, well, a tape drive. The only difference is, those moving parts are now Amazon's responsibility.

The old ways of using tape are dying, but tape itself will live on.

Tape is dead. Long live tape!