Exabyte

» see exabyte.com as it lived, on the wayback machine

~ the eulogy ~

The trick Exabyte was built on came from consumer video. Sony had spent years perfecting a rotating head that laid diagonal stripes across a narrow 8mm tape, because that is what a handheld camcorder needs to record an hour of television-quality picture on something you can hold. Juan Rodriguez and Harry Hinz left StorageTek in 1985 to point that mechanism at data instead of video, with Kelly Beavers as the third founder, and the company they started in Boulder shipped the EXB-8200 in 1987.

It held 2.3 gigabytes on a cassette the size of a pack of cards, at a moment when a large hard disk held under a hundred megabytes. The drive cost about two thousand dollars and the media cost almost nothing, because it was a camcorder tape and the world was already producing those by the million. Sun shipped Exabyte drives in its workstations. Unix administrators ordered them without comparing anything, because there was nothing at that capacity to compare against.

The numbers got large quickly. Revenue went from $89 million in 1989 to $287 million in 1992, Fortune ranked the company third among the fastest-growing in America in October 1991, and by 1994 around 650,000 of the drives were installed worldwide. Then the successor product slipped. Mammoth, promised at 40 gigabytes, arrived in March 1996 two years late, and Quantum's DLT 7000 spent those two years taking the accounts. Exabyte never recovered the position, merged with Ecrix in 2001 to get the VXA format, cut 250 jobs in March 2001, put itself up for sale on 30 June 2006, and closed as a company on 20 November 2006 when Tandberg Data completed a $28 million purchase of the assets.

Backup where I was, around 2000, meant a spindle of blank CD-Rs and a burn that failed often enough that you checked it afterward. Nobody I knew owned a tape drive, because tape was never a domestic object here or anywhere else. It lived in rooms with raised floors and a service contract. That is the honest distance on this one: Exabyte was one of the most important products of its decade and almost nobody outside a machine room ever laid hands on it.

~ the dossier ~

~ the origin ~

Helical scan is a strange idea if you come to it from data storage. Ordinary tape drives write straight lines down the length of the tape, which is simple and slow. Helical scan spins the head at an angle so the stripes run diagonally, packing far more into the same length of tape at the cost of a mechanism with moving parts under tension. Video engineers had accepted that trade decades earlier because there was no alternative for a picture. Exabyte's contribution was noticing that data would take the same trade.

Between 1991 and 1993 the 8mm products were about eighty percent of the company's revenue, which is a good position and a dangerous one. The format was proprietary and the customer base had been won on a capacity lead rather than on any lock-in. Both of those conditions were fine while the lead held and worthless the moment it went.

~ the unraveling ~

Mammoth was supposed to extend the lead and instead demonstrated how quickly it could be lost. The drive was technically the better product when it finally shipped in March 1996, and it did not matter, because the buyers who needed a decision in 1994 had already made one. Quantum's DLT was linear rather than helical, a less clever mechanism that tolerated dirty environments better and could be bought that year.

What finished the category was not a single competitor. HP and IBM put their own formats aside, brought Seagate in, and agreed on LTO as an open standard with multiple licensed manufacturers, which meant a customer buying tape no longer had to bet on the survival of one company in Colorado. Exabyte was, by the end, a company selling a proprietary format against an industry consortium, and the outcome of that arrangement is never in doubt.

~ what got buried with it ~

The Ecrix merger in 2001 is worth a line because it was almost a good ending. Ecrix had been founded in 1996 by two of the original Exabyte founders, and VXA wrote data in packets so a drive could reread a bad section instead of failing the whole restore. That is a genuinely better idea about what backup is for. It arrived attached to a company that no longer had the balance sheet to push it.

Tape itself never died. It sits in the deepest tier of every large archive and LTO is still shipping new generations, because the economics that made an 8mm cassette attractive in 1987 hold up just as well at petabyte scale. What died was the notion that a storage format could be one clever company's property. Exabyte proved the idea worked and then got beaten by everybody agreeing on a worse one together.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »