Tape doesn’t sound like a technology to help shape business computing in the early 21st Century, but its success continues to confound those who would rather it shuffled off its reel-to-reel in favour of more exciting designs such as optical media, hard disks or even, improbably, solid state. Its very longevity (anyone remember using 7-track?) is the root of its unfashionable image; a technology used as long ago as the dark ages (circa 1972) must, surely, be begging for a trip to the skip.
Tape survives because it has evolved a simple survival strategy. Just when you think it is heading out the door, engineers come up with another innovation in capacity and performance to appease the gods of data. The biggest jumps in tape systems are always taken when a technology vendor decides to bury backwards compatibility as a way of making possible large increases in capacity.
The best recent example of this is Sony, which has recently started shipping the latest version of its AIT technology, now knows as SAIT-1 (Super Advanced Intelligent Tape), which packs an extraordinary 500 GB (>1 TB compressed, the first drive to manage this) onto a cartridge with a data transfer rate of up to 30 Mbyte/s (see table for comparison with AIT-1/2/3) when hitched to either Ultra 160 Wide SCSI or Fibre Channel interfaces.
It ditches AIT’s 8mm tape format - outwardly identical in format to 8mm camcorder media - in favour of a half-inch single-reel cartridge, so goodbye to backwards compatibility, but the migration path for the new design is claimed by Sony to keep tape ticking over until past the end of the current decade.
Beyond the Gigabyte
This is amazing stuff. A technology that should have been dead and buried by now, has worked out how to keep itself current for the best part of another decade by proposing single cartridges with multi-Terabyte capacity. It would not be a surprise if the engineers turn round in 2010 and announce a new tape format capable of capable of offering three figure uncompressed Terabyte volumes on a single cartridge – and so the cycle will continue.
So how are the boffins doing it? More to the point, why are they doing it?
SAIT-1 uses a recording design known as ‘helical scan’, identical in principle to the way a common VHS video player works, and in fact the design originated in TV recording systems of the 1960s. Instead of recording onto tape that is moving across a static read/write head at great speed (as in the rival, ‘linear’ scan), in helical recording the tape moves very slowly over a rapidly spinning drum head. People argue endlessly about the pros and cons of helical versus linear tape mechanisms (some have accused helical designs of having a shorter read/write life) but with the arrival of SAIT and its huge data capacities and rapid data retrieval, it is possible that helical scan will start to win out.
Multiple read/write heads are claimed to increase data transfer rates (hence the 30MB/s figure), while more rapid tape positioning results from lower tape velocity onto the spool (no acceleration and deceleration times as with linear tape). Sony makes great play of the recording density of its AME (Advanced Metal Evaporated) media, though it is hard to assess these claims against the manufacturing processes of rival tapes.
More importantly, the data capacities possible with helical scan are several times what is possible with equivalent linear designs. If you’re running an archiving system then the more capacity you can fit on a single cartridge the better; fewer cartridges reduces media load and unload times and conserves physical space. You might even get away with fewer backup units and find that the greater capacity means that more data can be ‘online’ at one time relative to the number of drives.
However boring and dependable it has been in the past, the key driver for tape storage technology is now very clearly the explosion in hard disk capacity. Very simply, as drive capacity has increased at the same time as plunging prices have made large drives more widespread, backup systems have had to keep pace. Hierarchical Storage Management (HSM) and compression have done some of the work, but the arrival of high-capacity SAIT-1 has still plugged an important hole in the capacity story. In an age where large companies store Terabytes or even Petabytes of data offline, tape has had to be able to handle the requirement or risk seeing its star wane.
The economist Keynes famously and pithily noted that ‘in the long run we are all dead’. In other words, the problems of tomorrow are not something we necessarily want to bother about today, so people tend not to. This provides a parallel of sorts for the sales philosophy of data storage companies. They all quote long MTBFs for the drives, and 30-50 years shelf life for the tapes themselves, but everybody knows that no storage medium hangs around that long. People upgrade – indeed are made to upgrade as technology improves. What matters in the short run is having technical credibility when making the sales call.
Eccentrically, however, some organisations do hang onto data for decades, leaving it mouldering in back rooms somewhere in case it turns out to be needed beyond its immediate context. Public bodies, government, scientific institutions and, you presume, universities, are the extreme examples of this tendency to squirrel old data, but commercial enterprises are also increasingly prone to data retention. The problem with tape medium longevity – and this will apply to SAIT-1 as much as any other design in tape’s history - is that preserving the media is only half the game. You also need a working drive to read it on, and that is usually a problem.
Apparently NASA still has research data from the Apollo moon program held on 7-track tape that it can only get to on a few rickety old drives. Let’s hope they hold out, or some of Harrison Schmitt’s famous and scientifically productive jaunts around the moon during Apollo 17 will be little more than history – literally.
One capability that Sony will get round to adding to S-AIT at the end of the year is Write Once, Read Many, or WORM for short. Another first for tape systems, this will make it possible to write data to media with a guarantee it can’t be subsequently over-written. This is becoming more of an issue for companies operating under ever-stricter financial services regulation, where data-time integrity is essential for record keeping, and is likely to spread into related fields as time passes.
Longer term, tape streamers are set to be an integral part of companies networks, and it is looking increasingly likely that the technology is undergoing a renaissance that will take its usable life expectancy well into and perhaps even beyond the second decade of this Century. If it makes it to 2020, that will make it the longest serving ‘perishable’ technology in the history of computing.
Third-party SAIT products will be available this summer from Qualstar and Grau, and not forgetting Sony itself.