IBM 350 Disk Stack

1956 · Disk Storage · Work 615 of 1,871

IBM 350 Disk Stack, 1956, Disk Storage

Fifty large polished aluminium platters stacked on a common vertical spindle with even gaps between them, the edges catching light in a repeating stripe of bright and dark bands, the topmost platter's surface showing faint concentric traces, the platters coated a dull magnetic brown on their faces, the spindle of bare turned steel, seen close from the side so the stack fills the frame edge to edge, soft light from behind passing between the platters and laying a bright line along each gap, the stack alone against pure black.

What it was used for
The disk stack of the IBM 350 was the bulk store of the IBM 305 computing system: fifty aluminium disks 24 inches across, painted with iron oxide, on a spindle turning at 1,200 revolutions per minute. Both sides of every disk were written, a hundred tracks per side, 2,000 bits per square inch, five million characters in all. Average access time was 600 milliseconds. The unit stood five feet high and six feet wide and weighed, by Tom Gardner's account, 1,730 pounds, plus a separate air compressor of 441 pounds. It was meant for the running bookkeeping of a business: accounting and inventory control, where any single record may be needed at once.
What was new
What was new was that order no longer mattered. Punched cards and tape have to be run through; here, in Albert Hoagland's account, any record spread over 240 square feet of surface could be read or written in under a second. That made bookkeeping reversible: a figure could be changed without re-sorting the file. Two catches. First, the famous five megabytes: these were five million six-bit characters, so 3.75 megabytes formatted - Gardner traces the widespread five megabytes to confusion with the eight-bit characters of modern drives. Second, the unit was not cheap but also not the expensive part of the installation: the 305 system rented for $3,200 a month, of which the historian Emerson Pugh attributed $650 to the disk unit.

What the sources leave open: The dates are of different kinds: Tom Gardner reports an internal IBM announcement on 4 September 1956, a press conference on 14 September 1956 and inconsistent statements in IBM's own archives; the first pre-production unit went to Zellerbach Paper in June 1956, the first production shipment only in November 1957. The ASME landmark page gives 100 tracks per side and at the same time 20,000 tracks in total; the two do not agree.

Further reading