Logo IMG


Terabyte Territory

Brian Hayes

My Cup Runneth Under

Something more than ongoing technological progress is needed to make multiterabyte disks a reality. We also need the data to fill them.

A few people and organizations already have a demonstrated need for such colossal storage capacity. Several experiments in physics, astronomy and the earth sciences will generate petabytes of data in the next few years, and so will some businesses. But these are not mass markets. The economics of disk-drive manufacturing require selling disks by the hundred million, and that can happen only if everybody wants one.

Suppose I could reach into the future and hand you a 120-terabyte drive right now. What would you put on it? You might start by copying over everything on your present disk—all the software and documents you've been accumulating over the years—your digital universe. Okay. Now what will you do with the other 119.9 terabytes?

A cynic's retort might be that installing the 2012 edition of Microsoft Windows will take care of the rest, but I don't believe it's true. "Software bloat" has reached impressive proportions, but it still lags far behind the recent growth rate in disk capacity. Operating systems and other software will occupy only a tiny corner of the disk drive. If the rest of the space is to be filled, it will have to be with data rather than programs.

One certainty is that you will not fill the void with personal jottings or reading matter. In round numbers, a book is a megabyte. If you read one book a day, every day of your life, for 80 years, your personal library will amount to less than 30 gigabytes, which still leaves you with more than 119 terabytes of empty space. To fill any appreciable fraction of the drive with text, you'll need to acquire a major research library. The Library of Congress would be a good candidate. It is said to hold 24 million volumes, which would take up a fifth of your disk (or even more if you choose a fancier format than plain text).

Other kinds of information are bulkier than text. A picture, for example, is worth much more than a thousand words; for high-resolution images a round-number allocation might be 10 megabytes each. How many such pictures can a person look at in a lifetime? I can only guess, but 100 images a day certainly ought to be enough for a family album. After 80 years, that collection of snapshots would add up to 30 terabytes.

What about music? MP3 audio files run a megabyte a minute, more or less. At that rate, a lifetime of listening—24 hours a day, 7 days a week for 80 years—would consume 42 terabytes of disk space.

The one kind of content that might possibly overflow a 120-terabyte disk is video. In the format used on DVDs, the data rate is about 2 gigabytes per hour. Thus the 120-terabyte disk will hold some 60,000 hours worth of movies; if you want to watch them all day and all night without a break for popcorn, they will last somewhat less than seven years. (For a full lifetime of video, you'll have to wait for the petabyte drive.)

The fact that video consumes so much more storage volume than other media suggests that the true future of the disk drive may lie not in the computer but in the TiVo box and other appliances that plug into the TV. Or maybe the destiny of the computer itself is to become such a "digital hub" (as Steve Jobs describes it). Thus all the elegant science and engineering of the disk drive—the aerodynamic heads, the magnetoresistive sensors, the ruthenium film—has its ultimate fulfillment in replaying soap operas and old Star Trek episodes.

David Thompson, now retired from IBM, offers a more personal vision of the disk drive as video appurtenance. With cameras mounted on eyeglass frames, he suggests, we can document every moment of our lives and create a second-by-second digital diary. "There won't be any reason ever to forget anything anymore," he says. Vannevar Bush had a similar idea 50 years ago, though in that era the promising storage medium was microfilm rather than magnetic disks.

comments powered by Disqus


Of Possible Interest

Letters to the Editors: The Truth about Models

Spotlight: Briefings

Computing Science: Belles lettres Meets Big Data


Foreign-Language PDFs


Subscribe to American Scientist