This article is more than 1 year old

Flash, holograms? Where will your archive end up?

Will you still tape me tomorrow?

Setting tape free

“The other key factor is LTFS, which lets you see tape as NAS and is also portable.” He adds that a vital part of implementing an archive today is to “use open and/standard formats as much as possible, so that's REST, LTFS and so on.”

LTFS was developed by IBM before being passed to SNIA (the Storage Networking Industry Association) for future development, and as its name suggests it is a file system for LTO. The big benefit over previous tape indexing schemes is that LTFS holds the metadata in a standard format and on the tape itself, instead of in a proprietary format and a separate database. That means an LTFS tape can contain a hierarchical directory structure and be read on any system. Similarly, REST is a software architecture style that allows (among other things) components and messages to be self-describing.

The shift to LTFS is already under way in industry sectors that have long experience of archiving, such as the broadcast industry, says Walter Fusi. “The media & entertainment segment is well used to archiving, more so than newer ones such as surveillance or HPC.,” he notes. “The big change for them is a move from proprietary solutions from Sony, Panasonic and so on to standard ones.

“They're heavy users of tape because there's no other option for such large volumes, so the answer came from IBM, with LTFS. It's the first opportunity to write tapes with a standard file system, readable in any LTO drive.” He continues: “People tend to think of tape as old and slow. Tape is not slow! Modern tape is extremely fast when you're working with big files.”

“Tape still has a long way to go – I've been to IBM's labs, there is 100TB tape on the way,” agrees Matthew Addis. He adds that higher tape densities and read/write speeds are great for cloud archive providers such as his company, because it means their running costs should fall over time.

Other technologies are also evolving, though. As Matt Starr puts it, “Can flash get cheap enough to replace rotating disks? Possibly. Can disks get cheap enough to replace tape? Also possibly.” He notes that, like many other storage developers, Spectra Logic already uses flash buffers to accelerate its archiving systems. “Today it is flash for the metadata, an enterprise SAS landing zone for the data, and then tape, but we will see a disk tier also coming in for near-line,” he says.

Return of the disc

Optical storage could be due a come-back, too. It was once a big player in near-line archiving, first with 12-inch write-once discs, then 5.25-inch magneto-optical, and more recently with CD and DVD jukeboxes. Optical lost out, in part to cheaper hard disks, but Walter Fusi says it is now regaining favour.

“We've recently seen the return of optical devices from two key players,” he explains. “People misunderstood optical, so they bought a DVD jukebox but then filled it with cheap media which created frequent data loss and a loss of confidence. But in Japan they never gave up, so now Sony and Panasonic are back with a safer, more user-friendly Blu-ray. They control the media quality and put up to 15 discs in a box the size of an LTO cartridge, so that's 1.5TB, and with 30GB discs promised that's 4.5TB per cartridge.”

He adds that with good media, optical could save on tape conditioning, plus it is not affected by magnetism, is more tolerant of temperature, and is better for random access. “It won't replace tape or hard disk, but it will win share,” he says

A somewhat longer bet is holographic storage – in effect a three-dimensional version of optical storage, offering perhaps 200 times the capacity of Blu-ray by storing data as images in a 1.5mm-thick 3D layer, rather than as bits on a 2D surface. Dr Ken Anderson, the CEO of holostorage pioneer Akonia, claims that it will cost less than 25 US cents per terabyte by 2020.

Of course, no-one can be sure what the likes of LTO will cost by then. However, today's proven LTO-6 is well under $10 per terabyte, and extrapolating from older generations gives us LTO-9 around 2020, with 10 times the cartridge capacity of LTO-6. Maybe 50 cents per terabyte?

There are just too many 'maybes' here for Matt Starr. “I don't see other technologies on the horizon,” he says. “Holographic has been around since the 1990s in the labs, it's a great idea but getting it into a product is the tricky part. BluRay's price per platter is too expensive for now.”

His tip for that near-line tier is hard disks using high-density shingled magnetic recording, or SMR, which he says is ideal for applications where the access pattern is read-mostly with few edits. The editing issue is that SMR overlaps the bits as they are written on the disk medium; this gives higher data density but means you can't rewrite a bit in the middle of a used track, instead you have to reclaim the space and rewrite the data.

“Tape will be the primary player, from a cost and power point of view, but as flash encroaches on disk, disk moves down the food chain. We are very impressed with SMR drives, it will be a very viable archiving product in the short term,” he concludes. ®

More about

TIP US OFF

Send us news


Other stories you might like