Scene: The ancient Sumerian philosopher lays down his stylus one final time, reflecting upon the newly completed set of clay tablets--his life's work. That night, a rampaging dromedary in the temple precincts will crush the tablets, grinding them back into the desert clay from which they were formed.
Scene: Renowned for his lifelike depictions of jousting tournaments, the painter adds his signature to the room-sized fresco in the royal dining hall. The artist is unaware, however, that the impurities in his pigments will cause his masterwork to disintegrate, flake by flake, over the next six months.
Scene: The editor nudges the cut one frame this way, two frames that way--perfecting the timing of the program's pivotal scene. She has rendered the composites and matched them back into the timeline; the scored and sweetened soundtrack is laid back in perfect sync.
Then, without a moment's warning, the RAID array unmounts itself, the monitor goes dark, the speakers silent, and all that remains is the rhythmic tick-tick-tick of a hard drive's head actuator, methodically scraping the pictures off the platter.
SWEARING FOUL OATHS
It's impossible to estimate the number of times each year this sad tale plays itself out in edit suites across the world. I've witnessed two big-time failures this past year in our own shop, and each has been a horrible experience, beyond simple repair costs. Weeks worth of work was lost in each case, along with the good will of clients whose jobs were trashed.
Particularly ironic is the notion that this most crucial part of the digital content chain--storage--is left to the one remaining mechanical device in a modern computer, the hard drive. Even within the relatively recent time span of desktop video, futurists and prognosticators have foretold a time when motors and magnets no longer have a role in moving pictures. And that time was supposed to have been now.
I recall one such ingenious scheme, described about 10 years ago, called "wafer scale integration" (WSI).
When RAM chips are manufactured, it seems, a large number of junctions are deposited onto a single silicon wafer, and later cut into individual RAM circuits. In WSI, a single silicon wafer full of RAM, perhaps the size of a CD, would be left intact, with a common path leading to and from the circuits, and with a small communications circuit placed ahead of each chip. A storage controller asks permission to store a byte; the first RAM location is queried, and if the location is reported as empty and viable, a byte is written--no moving parts.
SEARCHING FOR STORAGE
I'm not sure why WSI never succeeded, although it's not uncommon for brilliant ideas to eventually reveal fatal flaws. Of course, innovators don't stop inventing after one area of investigation goes sour, and the storage we all use for video and audio seems to constantly evolve. I spoke recently to Michael Anderson, chief engineer for array manufacturer Huge Systems (now part of Ciprico).
Mike was a founder of Huge Systems, and before that, of Medea, both of which specialize in creating massive high-speed arrays from inexpensive ATA drives. Older arrays exclusively used pricey, and often smaller, SCSI drives. Mike shared the story of a talk he gave at a major post-production conference during which he asked the audience of several hundred editors to raise their hands if they used SCSI storage.
According to Mike, about 5 percent of the audience responded. The balance, when asked, indicated they were using ATA arrays. The response remains a surprising indication of the rapid time-to-adoption for ATA-based devices.
It's normal for the cost of technology to drop as a product life-cycle progresses, but ATA arrays leapfrogged the price curve as drive capacities increased and parts decreased. All I know is that we bought less than 50 GB of dual-channel SCSI RAID storage back when we got our first nonlinear editor, and it cost more than $10,000. Our new 1.2 TB ATA-based RAID array, with dual-channel ultra 320 SCSI interface, cost less than $5,000.
Lower cost and higher speed are attractive features, to be sure, but the fundamental flaw remains--electrical and mechanical breakdowns can spell catastrophe. And it may be that this moment in history offers fewer clear choices for prevention and remediation.
Larger RAID array capacities have made it more practical than ever to use "mirroring"--using half the array to make a simultaneous backup of the data you write.
In practice, though, it's not foolproof. Hardware damage can just as easily affect the backup, and retrieval of the mirrored data is no mean feat.
And as for backups onto other media... well, the practice hasn't advanced much in all these years. Tape backup drives are profoundly expensive. The tape itself isn't cheap, and most data formats still run abysmally slow, far slower than real-time dubs of the actual footage onto less expensive digital video media. What's an editor to do?
One candidate technology may be Tapestry, a system of holographic archival data storage media developed by InPhase Technologies. The technology grew out of Bell Labs research, and with a team of former Lucent and StorageTek execs at the helm, it has garnered support from Maxell and other industry leaders. The physically compact Tapestry media is planned to hold up to 200 GB with a shelf life of more than 50 years. Now, that's what I call archival. We'll all get a good look at Tapestry when it debuts in 2006.
Until then, save early, save often, and keep your fingers crossed.
TELEVISION TAPE UPDATE
In January, I told you the story of a television pioneer, John Vrba, who helped make a case for the use of a then-new invention, videotape, for commercial production. I invited readers to download and view John's 10-minute kinescope pitch, produced by KTTV in 1962, from a Web site I'd thrown up for that purpose, http://www.televisiontape.tv
I thought you might be interested to know that the "Television Tape" site has been seen by more than 2,500 unique visitors from 25 countries. Who knew that the good people of Togo, Nigeria and Lithuania would be interested? Among the IP addresses have been lots of broadcast stations and post-production houses. Military installations have been curious, as have the media departments at colleges and universities. A link at a film archivists' bulletin board drew lots of viewers. I also noticed a visit from the Museum of Television and Radio--interested, folks? In all, the 5 MB movie racked up about 30 GB in downloads, far more than I'd expected, and that's great. Thanks to all who posted on the guest book or e-mailed kind words.
Walter Schoenknecht can be reached via e-mail at walter@mmgi.tv.
