Why The CD Tower Crumbled - the arrival of DVD-ROM - Technology Information

Computer Technology Review, Sept, 1999 by Hal Glatzer

As towers go, it wasn't very impressive. The Eiffel Tower's taller; the Tower of Babel's more fabled; and the Leaning Tower of Pisa ...well, let's just say it sustains greater angular momentum.

But what about the CD tower? Only a few years ago, a rack of six or seven CD-ROM drives, each spinning a single platter, was the epitome of CD storage. With no robotics to impose delays and every disk sitting right under a read-head, everything in a tower was on-call all the time. It offered practically instantaneous retrieval of any file from enormous databases.

The killer app for CD towers was the otherwise humble telephone directory. Database engines accessing a six-disk tower had more than 3GB of data at their disposal. Users could type in a name and the software could display that person's (listed) phone number. With almost no extra effort, it could perform cross-directory searches, too, such as finding all the phone numbers installed at a given address or generate a selected list of numbers based-on specific criteria.

Of course, towers weren't cheap: their cost being a multiple of the price per drive at a time (the early 1990s) when CD-ROM drives cost a few hundred dollars apiece. But there were some tradeoffs that made towers cost-advantageous. Unlike jukeboxes, towers had no other moving parts, so support expenses were negligible. Other than time spent replacing disks (which, for a phone-directory, might happen only once a year), there was hardly ever a need to take a drive off-line. Almost every tower enabled administrators to do what's now called "hot-swapping" to replace any drive in the tower without shutting down or interrupting the activity of the others.

If there was ever such a thing as a "passive" storage peripheral, the CD tower was it. You set it up and you left it alone.

A CD jukebox, by comparison, with its racks, gears, pickers, and doors is a Rube Goldberg invention that can't be left alone for long. (I'm not talking about reliability and MTBF statistics-- I mean simply that managing a jukebox is an ongoing process requiring human attention; someone always has to add or remove disks, or check to see why some disks are more frequently-requested than others, etc.)

Jukeboxes proved to be a better investment than towers, though, because they could hold ten or 20 or even 100 times more disks than towers could. Where towers did beat jukeboxes was in retrieval times. Even a 50-disk library will force users to wait in a file-retrieval queue that can sometimes seem interminable. For administrators and managers, the high capacity of jukeboxes outweighed the faster retrieval time of towers. The loss of productive time while users were being inconvenienced was less important than obtaining storage at a lower cost per megabyte. Jukeboxes were simply a better investment.

But what users really want (as if you didn't know) are both: rapid access and high capacity. And that came about with the addition of a hard disk drive to the jukebox's front end. Once that happened, the tower was doomed. Even the earliest configurations at a time when HDDs cost more than a dime-a-megabyte were more cost-effective than CD towers could ever be. As the price of hard disk storage trended downward, RAIDs became even more attractive than single HDDs. Today's RAIDs can easily hold 10-20 percent of a jukebox's capacity for only about a nickel-a-megabyte.

Interestingly, the "SolidState Disk" (SSD) hasn't risen to the occasion, except in some key applications. SSDs are much faster than HDDs; so, theoretically, they should be more attractive. But SSDs cost far more: about $5/MB. For the steady or predictable retrieval of large or rarely modified files such as document images, HDDs are certainly fast enough. SSDs, however, are better suited to tasks that require manipulating and/or updating many small files almost simultaneously, which is what happens in applications such as electronic funds-transfer and the processing of other commercial transactions. The cost of SSDs is falling, to be sure; it's likely that jukeboxes optimized for video and multimedia applications may have SSD caches, soon, rather than HDD caches. But SSDs will always be more expensive than HDDs. By the time SSDs approach $2/MB, HDDs will cost only a few pennies per megabyte.

The arrival of DVD-ROM has also doomed the tower. The same telephone directory that previously needed three or four CDs to hold it can now fit easily on a single 4.7GB DVD-ROM disk; and read-only DVD-ROM drives, which cost only about $100 more than CD-ROM drives, can also read CD-ROMs and CD-Rs and CD-RWs.

Installing DVD-ROM readers in a jukebox adds relatively little to the cost. Installing DVD-RAM drives (or DVD+RW, when they arrive), adds record-ability and rewritability for a few hundred dollars more per drive. But recordable DVD drives also read all the replicated DVD formats and all the replicated and recordable CD formats, as well. One size really does fit all.

So the tower, attractive as it once was, is now only an empty shell: a stepping-stone on the path toward... oh, to heck with the metaphors! You will never see--or need--a "DVD tower."