In 1956, the day after the IBM ® 305 RAMAC computer was announced, The San Jose Mercury News ran a story under the headline, “A machine with super memory!” The story noted: “The information on the discs can be added to, altered or erased at will. Card-sorting, one of the most time consuming office-machine processes, is eliminated or greatly reduced.”
RAMAC was the first computer to use a random-access disk drive—the 350 Disk Storage Unit—the progenitor of every hard disk drive made in the 55 years since. But the impact of the 350 Disk Storage Unit was more than that: It introduced the concept of instantly accessible information. Before RAMAC, information had to be entered by running a stack of cards through a punched card machine, and answers would arrive in hours or days. RAMAC could find data in seconds, alter it, and move on to find a completely different piece of data. It let enterprises think about data in new ways, mixing and matching it on the fly. Random access made the relational database possible.
The idea of some kind of magnetic, quickly accessed memory had been around since the late 1940s. Small companies like Engineering Research Associates of St. Paul, Minnesota, developed magnetic drum storage. These rotating drums were reliable but slow. An Wang, who built the word processing giant Wang Laboratories, developed an early technology that made magnetic core memory possible. Those devices of wires and tiny magnets had quick access times but couldn’t hold much information. They couldn’t replace tape or punched cards, but were used as short-term memory for early computers—the predecessors to the solid-state memory of later DRAM chips.
In 1952, with computer excitement in the air, IBM sent Reynold Johnson to San Jose to start a new research lab. When the Air Force wanted a random access inventory system, Johnson set his 50-person lab in motion, trying everything—strips, rods, tapes, flat plates, you name it. In the early 1950s, no one had any idea how to make a fast, reliable random access memory machine.
The technical obstacles were enormous. The lab soon settled on using spinning, horizontal disks coated with magnetic material. Magnetic spots on the disk would represent a character of data. Since the spots would have a magnetic field, a magnetic arm, like a record needle, could hover over the spots and read them as the disks rotated at high speed. But the first challenge was finding material for the disks. The disks had to be perfectly flat, strong and light enough to be spun by a reasonable-size electric motor. A single aluminum disk warped at high speeds. After much trial-and-error, the researchers tried gluing two aluminum disks together. It worked.
Even more problematic was the arm. It could never touch the disk or it would wipe out the data. Two researchers, William Goddard and John Lynott, came up with an arm that fired out compressed air to hover just above the disk. “When we concluded we could do that, we could see a fairly clear road to building a practical random access memory,” said Louis Stevens, a senior engineer at the San Jose lab.
Since a single disk could not store enough data to be useful, the researchers built a machine that held 16 disks stacked horizontally, with a tiny space between each disk. “We called it a jukebox and a meat cutter and a lot of other things,” Johnson said. To disbelieving executives at headquarters, it was known as the baloney slicer.
The jukebox set-up left the San Jose team with one other challenge: how to get that arm to the right place on the right disk in the blink of an eye. Said Johnson: “That was our goal, to go from any track which was six inches in on a disk, out, down, two feet to the bottom, and in six inches—in half a second. We achieved something like 800 milliseconds and that’s where the product came out. It was really an amazing achievement.”
The product as it was introduced was the size of two kitchen refrigerators side by side. Fifty disks inside spun at 1200 revolutions per minute, while the arm dashed in and out, accessing the data at about 100,000 bits per second. The whole thing could store 5 million binary decimal encoded characters at 7 bits per character—in other words, about the contents of a Manhattan, New York, phone book. Each character could be read or changed randomly, in any order, at any time, making the machine a revelation. Hugo Cannizzaro was selling out of IBM’s San Francisco office at the time. “It was the first time I had customers insisting that I come to their offices immediately so they could place an order,’’ he said.
At the time, thanks to the success of the early Univac computer, the suffix “AC” was a popular tech branding them—like the prefix “e-”in the dot-com era. The researchers called their computer Random Access Memory-AC, or RAMAC. The IBM marketers liked RAMAC, but changed it to mean Random Access Method of Accounting and Control. The 350 Disk Storage Unit was the key feature of a whole accounting system. More than 1000 RAMAC 305 systems were built, but the number belies the effect on the market. RAMAC was the beginning of the end of ubiquitous punched cards, and the start of a nearly six-decade drive for real-time information.